The Use of Lasers for Pavement Crack Detection
Contents
1. 9 89 on viii 14 15 16 19 21 22 23 25 27 27 39 40 LIST OF FIGURES o Wer a a a 183 6 Patching e Aa a a Tey d la J r oa 1326 730 Alligator cracking KK 9 BLOCK Cracking 5 o A A ola Transverse cracking o Longitudinal cracking Laser probe and laser calibration board Laser orientations 4 4 e e e e s x 8 Calibration board results Laser probe and probe processing unit Pulsed modulated infrared light from GaAs lasers ou 97 5 2e de ee e Triangulation principle Laser measurement range K CPU sub rack with power supply and receiver averaging boards Data acquisition DAQ boards Crack system in the profilometer Power spectral density plots of different eracking types 4 2 Geo a Ue X c ue Power spectral density plots of different Cracking severity e K e wo o ix 42 43 45 47 48 Running mean slope threshold technique ADE Ten to moderate alligator cracking data e Running mean slope threshold technique applied to severe alligator cracking data Raw data Filtered data with r 0 SEE value SP every 16 data points Actual r 0 and r 4 values for the data in Figure2. 3 313131 4 513151 da jar jer jin SERI 4 bat Teng ron 1506531 4 8 500 2335 MENATA le m m b s b b e e vi 57 x Y ELL rw in ant ent VET DR 8 OTTO TTY t UU 1 ll i _ i 1 azioni ui der f i al ite Try st 1 ERP pinnas ETT it En 135531 sin 6 St ae ME LL APPENDIX 8 RUNNING MEAN SLOPE THRESHOLD LISTING 59 60 Chkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkk This program inputs a filtered data file and detects cracks using the running mean slope threshold method described in Chapter V 0 0 0 0 0 0 6 FILEIN input data file FILEOUT output file containing crack info 6 for plotting 6 XM array of X values used in current mean XD array of X values used in lookahead if 6 values are decreasing array of X values used in lookahead if 6 values are increasing M pointer into XM array 6 D gt down counter 6 U up counter 6 DX value output to file for plotting cracks 6 200 ground level 6 300 above gnd level spikes or errors 100 crack 6 C counter of number of points in crack 6 if gt 96 6 will reset 6 Ic same as C but above gnd level 6 NPTS number of data points 6
3. This is done through a feedback mechanism The receiver averaging boards are located in the CPU sub rack as shown in Figure 3 5 There is one board for each laser probe Each board receives serial data from the gauge probe at a rate of 32 KHz and is capable of reducing the data rate by forming the average of a number of measurements The data rate also referred to as updating frequency is set by jumpers on the board The update frequency ranges from a maXimum of 32 KHz no averaging down to 62 5 Hz in powers of 21 31 25 usec AAA 16 pulses 16 pulses 16 pulses Figure 3 2 Pulsed modulated infrared light from GaAs lasers Triangulation principle 01 02 Figure 3 3 23 gt TT Yezda RESOLUTION IN A 13 98 MEASUREMENT RANGE IN GROUND LEVEL 10 08 0 10 VOLTS Figure 3 4 Laser measurement range two Output from the receiver averaging boards 35 the measured distance value represented as 12 bit parallel data plus a data invalid bit and a data ready flag This 12 bit parallel data value 15 input to the 68000 data acquisition board DAQ which interfaces to the COMPAQ s PC bus 3 2 68000 DAQ Board The data acquisition board initially used to determine the measuring characteristics and capabilities for the projeot is a specially designed board which uses the Motorola 68009 processor and plugs into one of the system expansion slots in the COMPAQ Portable III expansion
4. and evaluation as it had existing on board laser equipment The initial investigations proved that the lasers on the SDP could be used for crack detection Based on this result the study has proceeded in obtaining the necessary equipment and developing algorithms and software for implementing an automated crack measuring system which hopefully could be used to aid in PES This report discusses the first two phases of this project determining the feasibility of crack detection using the laser and obtaining and testing equipment so such a system could be implemented During these first two phases the capabilities and limitations have been identified To date it appears that a system can be developed with a limited capability for crack identification and reporting which could be useful for PES data collection activities The third phase of development and implementation of the automated crack identification system is currently in progress IMPLEMENTATION STATEMENT An automated and objective procedure for crack measurements and recording would provide a significant savings to the State during P E S data collection procedures It could be used in many other areas where statistical information regarding pavement cracking is desired vi 111 iv vi ix 11 11 11 12 13 17 18 18 24 24 TABLE OF CONTENTS PREFACE ABSTRACT SUMMARY IMPLEMENTATION STATEMENT s e LIST OF FIGURES CHAPTE
5. suggested not to change the F stop more than two settings They now recommend a setting of 2 8 CHAPTER III CRACK IDENTIFICATION SYSTEM HARDWARE The three basic hardware components of the initial configuration for the crack identification system are the optocator the 68000 DAQ board and the COMPAQ Portable III personal computer The optocator obtains a distance measurement using non contact lasers The 68000 data acquisition board acquires the data from the optocator at a specified sampling rate temporarily stores the data in onboard RAM and performs some preliminary processing of the data as well as data reduction Finally the COMPAQ accepts a reduced data set and stores it for final processing and analysis 3 1 Optocator The optocator is an optoelectronic measurement system which measures the distance to an object with high speed and precision Most importantly the measurement is made without contacting the measured surface The basic components of the optocator are the non contact laser probes the probe processing units PPU and the CPU sub rack which contains the power supply and the receiver averaging boards which receive and process data from the gauge probes A laser probe and probe processing unit are shown in Figure 3 1 The gauge probe contains a pulsed modulated 32KHz and intensity controlled laser diode a position sensitive photodetector and an appropriate lens system The laser diode is a class III b gallium
6. 5 2 FT Filtered Dota r 0 r 4 Figure 5 6 Filtered data with r 0 r 4 value plotted every 16 data points 48 Case III 2 Mousanda 6 o 0 Case III 8 r 0 r s Figure 5 7 Actual r 0 and r 4 values for the data in Figure 5 6 CHAPTER VI CONCLUSIONS AND FURTHER RESEARCH The report describes the first two phases of research Project 8 18 86 1141 for developing an automated method of obtaining and evaluating pavement distress and cracking information for PES For these initial two phases the use of two lasers one in each wheel path are used to obtain cracking data which is processed on a Motorola 68000 based data acquisition board and the COMPAQ Portable III For detailed analysis the data is filtered to remove the DC component and long wavelengths before processing The data is analyzed using several different statistical techniques Two techniques in particular have been shown to be very reliable These are the running mean slope threshold and the autocorrelation difference methods Software still needs to be written which take the results from these two methods and provide the detailed reporting that is the percentage and severity of each cracking type within the section Several important conclusions can be made as a result of this initial study First alligator and block pavement cracking can be detected using the Selcom lasers mounted in the wheel paths However it is unlikely th
7. Applications of Artificial Intelligence III SPIE vol 635 pp 564 574 1986 Mahler David S Final Design of Automated Pavement Crack Measurement Instrumentation from a Survey Vehicle Report No FHWA RD 85 077 Federal Highway Administration Washington D C 1985 Pavement Evaluation System Rater s Manual Texas State Department of Highways and Public Transportation 1987 65 1 2 3 4 2 6 7 8 66 Epps A A H Meyer I E Lattimore Jr and H L Jones Roadway Maintenance Evaluation User s Manual Research Report 151 2 Texas Transportation Institute Texas A amp M University September 1974 Smith Roger E Michael I Darter and Stanley M Herrin Highway Pavement Distress Identification Manual for Highway Condition and Quality of Highway Construction Survey Federal Highway Administration Washington D C 1979 Box George E P and Gwilym M Jenkins Time Series Analysis Forecasting and Control Holden Day Inc San Francisco California 1976 Jenkins Gwilym M and Donald G Watts Spectral Analysis and Its Applications Holden Day Inc San Francisco California 1968 Papoulis Athanasios Probability Random Variables and Stochastic Processes McGraw Hill Inc New York 1984 Oppenheim Alan V and Ronald W Schafer Digital Signal Processing Prentice Hall Inc Englewood Cliffs New Jersey 1975 Childers Donald G ed Modern Spectr
8. DDP The SDP is owned by the State and used for road profile measurements After experiments indicated that these probes could be used for such detection a system was developed to further study this capability and to determine how it could be used to implement an automated high speed crack identifi cation system The third stage is the implementation of such a system so its usefulness for P E S data collection activities can be deter mined This research report describes the first two phases of the research effort 17 Key Words 18 Distribution Statement No restrictions This document is available to the public through the National Technical Information Service Springfield Virginia 22161 Surface Dynamics Profilometer SDP Lasers Pavement Distress Measure ments Pavement Crack Identifica tion and Recording 21 No of Pages 22 Price 20 Security Classif of this page 19 Security Clossif of this report Unclassified Unclassified 78 Form DOT F 1700 7 8 72 Reproduction of completed page authorized THE USE OF LASERS FOR PAVEMENT CRACK DETECTION by Lynda Donnell Payne Roger S Walker The University of Texas at Arlington Research Report 1141 1 Crack Identification Using Lasers Research Froject 8 18 88 1141 conducted for Texas State Department of Highways and Public Transportation in cooperation with the U S Department of Transportation Federal Hig
9. and Stationarity The definitions of mean variance autocorrelation and autocovariance described above are based on statistical ensemble averaging That is they were based on observations at a particular time t In practice one does not have the luxury of an ensemble of waveforms from which to evaluate these statistical descriptors Typically these statistical 31 estimates are obtained from a single waveform x n by substituting time averages for ensemble averages Here x n represents a discrete time series For a stochastic process to be accurately described by time averages instead of ensemble averages the process must be ergodic Ergodicity requires a certain amount of stationarity that is the statistics must be independent of the time origin selected A random process is wide sense stationary if its mean is constant for all time indices and its autocorrelation depends only on the time index difference m where m n2 n1 The variable m denotes the time lag that is the number of time increments between time n2 and time nl All results reported in this study assume the data is wide sense stationary or at least locally stationary such that time averages can be substituted for ensemble averages 4 4 Statistical Estimates If a stochastic process is ergodic then E X1 E X2 E X3 E XN and the mean X can be estimated by N X 1 N E x n n 1 The autocorrelation and autocovariance functions no longer depen on
10. interface provides an 8 bit parallel interface for downloading large amounts of laser data to the COMPAQ 3 3 COMPAQ Portable III The COMPAQ Portable III is the user s interface to the entire system From the COMPAQ s keyboard the user can run diagnostic checks on the system collect a specified amount of 2H genes onu Figure 3 5 CPU sub rack with power supply and receiver averaging boards 26 data download crack data to the COMPAQ for storage and subsequent processing or enter a real time crack counting mode The programs which provide detailed crack identification and section analysis reside on the COMPAQ When the user runs a section of road to be analyzed the data is collected on the DAQ boards and then downloaded to the COMPAQ for off line analysis The real time crack count mode provides a rough estimate of the number of cracks seen as the van moves at highway speeds This estimate is performed by the DAQ board using a variance measure In this mode the COMPAQ is used to issue the command to the system and to display the crack count Figure 3 7 shows the system as it is currently running in the profilometer 27 Data acquisition DAQ boards Crack system in the profilometer Figure 3 6 Figure 3 7 CHAPTER IV TIME AND FREQUENCY ANALYSIS TECHNIQUES This chapter provides some of the basic concepts in the theory of time series analysis needed in the processing of crack d
11. module See Figure 3 6 Its function is to receive the laser data from the cptocator and perform some preliminary processing of the crack data before passing it on to the COMPAQ Portable III for final crack identification and section analysis The DAQ board is actually made up of two boards Schematics for the boards are included in Appendix A The main board contains the M68000 microprocessor static RAM EPROM serial and parallel I 0 and is capable of running independently of the other The second board is an auxiliary memory board which only contains buffers and an additional 512K of static RAM This board is used when large amounts of data needs to be stored in real time The main DAQ board features include an 8 MHz Motorola 68000 microprocessor 64K static RAM 64K EPROM two Motorola 68230 parallel interface and timer chips an Intel 8251 USART and the IBM PC interface The 8 MHz M68000 provides 500 nanosecond bus cycles The static RAM and EPROMs have 100 and 200 nanosecond access time respectively This allows memory reads and writes with no wait states THe M68230 PI T chips are programmed in the 16 bit port mode to provide the parallel interface for two lasers The timers on the M68230 provide interrupt signals at the required sampling rate The Intel 8251 USART gives an RS 232 compatible serial interface running at 9600 BAUD The serial interface is used for most of the communications between the DAQ and the COMPAQ The IBM PC
12. of cracking again shows more power at wavelengths less than 1 4 inch Slight alligator cracking was not included because it is not believed that the lasers are accurately measuring slight less than 1 8 inch cracks In summary the spectral analysis results indicate that data obtained from road pavements with no cracking is uncorrelated Data from pavement with cracking is correlated in fact the higher the degree and severity of cracking the more correlated the data is 5 4 Running Mean Slope Threshold Method The basic idea behind this method is that a running mean representing ground level is maintained and each new data value is compared with this mean to determine if it is a value taken from a crack or not The term running is used because the mean must be constantly updated using the new data points to maintain an accurate representation of ground level Data points which are determined to represent a crack or a surface too much above ground level perhaps an extraneous rock or spikes in the data do not contribute to the running mean calculation The running mean is an average calculated from the last N data points which have been determined to be at ground level N is user selectable typically 4 to 8 The simplest way to apply this technique is simply to compare each new data value to the running mean If it is below a threshold distance from ground level then identify it as a crack do not include it in the mean and advance to
13. onto a position sensitive detector in the laser probe allowing accurate distance measurements 5 Further explanation of the optocator and measuring principle will be provided in Chapter III Since the lasers were available highway department engineers wanted to know if these lasers could help identify pavement cracking The intent of this study was first to determine the feasibility of using the existing lasers of the SDP to identify pavement cracking Then if feasible the work would be extended to design and implement a system which would identify the specified cracking patterns Few operational systems for crack identification using laser probes have been reported in the literature Most studies for such systems have used video data 6 7 The research herein does not use an elaborate video camera system only existing lasers 1 3 Project Phases As noted above this study consisted of three phases and this report is concerned with the first two phases First the feasibility of using the existing laser probes on the SDP had to be investigated This involved determining whether or not the resolution of the laser probes was sufficient to detect cracking patterns Also the measurement update rate had to be considered to determine if the laser could supply the necessary sampling rate for crack identification at highway speeds Another item of interest was the real time issue That is how much if any of the processing and analysis c
14. the time index of the random variable only the time index difference The time index difference is referred to as the lag and denoted by m The autocorrelation r and the autocovariance c then become r m E x n m x n and c m E x n m X x n X r m x Assuming ergodicity the autocorrelation and autocovariance can be estimated by 32 N m r m 1 N m x n x ntm n 1 and N m c m 1 N m 2 x n x x n m X n 1 4 5 Power Spectrum Estimation Spectral analysis is any signal processing method that characterizes the frequency content of a measured signal In spectral analysis one is typically interested in obtaining a spectral plot which represents the distribution of signal strength at each frequency Peaks in the spectral plot show which frequencies are predominant in the signal Most power spectrum estimation is accomplished by either the autocorrelation or the direct method 15 The latter method has become the most popular because of the fast Fourier transform FFT algorithm developed in 1965 16 The FFT is a fast efficient algorithm for computing the discrete Fourier transform DFT of a time series The DFT determines a sampled periodogram in which the values of the periodogram for only a discrete number of equally spaced frequencies is computed rather than evaluating over the continuous range of frequencies The method for calculating the power spectra in this study was first propo
15. was that cracking of the same type and severity would show similar coefficients while the coefficients would be significantly different for a different type and or severity of cracking First several sections of test data were modeled to determine the number of coefficients to use It was found that only the first three coefficients contributed significantly that is beyond three lags the coefficients were essentially zero This was also substantiated by the fact that the variance of the white noise the error term could only be decreased to a certain level by adding AR terms beyond that it really did not improve the model by adding additional terms Having determined that three terms should be used in the model different types of cracking were then examined Blocks of data one foot in length were examined It was found that data with more cracking showed higher autocorrelation values and the coefficients were significantly larger than data with no cracking However the resolution required to provide the detailed information needed simply was not there This technique could tell if there was a large amount of cracking or little to no cracking in each one foot block but that was all Since the details such as approximate number of cracks or severity could not be ascertained this method was not considered further 47 09 Case 111 0 8 0 7 0 6 0 5 0 4 Mousonds 0 5 Case IV 0 1 2 1000 DATA POINTS
16. 5 6 CHAPTER 1 INTRODUCTION 1 1 Project and Report Scope This project was initiated to determine the feasibility of using the laser probes on the Surface Dynamics Profilometer SDP owned by the the State Department of Highways and Public Transportation SDHPT for crack detection and identification found feasible a system was then to be developed for use on the ARAN measurement vehicle also owned by the State The SDP was selected for the initial testing and evaluation as it had existing on board laser equipment As will be discussed initial evaluations proved that the lasers on the SDP could be used for crack detection Based on this result the study has proceeded in obtaining the necessary equipment and developing algorithms and software for implementing an automated crack measuring system for PES This report discusses the first two phases of this project determining the feasibility of crack detection using the laser and obtaining and testing equipment so such a system could be implemented The third phase of development and implementation of the automated crack identification system is currently in progress and will be reported on in a later report This introductory chapter will first provide a background and general understanding of the crack detection and identification problem Further it explains some necessary terms and describes the project requirements Chapter two then addresses the feasibility issue It de
17. CS 52 an v An AL EMI x 51 Aa 3 C LES 491 5 33 F4 ti 5 374 an Eb era uur EE ANAIS CTE 2 3 i 5 1 sed iP x 3 180 08 CUG 9 9 94 141 1 cen 420514 I Enika mn 3 8 1 D Ul DNI EE 34 FN 1 DI rf e PEPE Ad de AL u THER 3 E AS i RESETE rt dk CATE he We POETS ae Pik D Pr7 118 8 RS Arca Pre Pt ac H 24 PCS PL 9 EL TA Toute BETE TR STACK EE AL 18 8 RER TALK Cle 94 Balas G4c08 2U0 8900 T 1eW lf 1 aer ir 41 rar zur PERI E TEXTS l rwr T sth 42 47 4 Wer uv gt 4 Paar lei n ar 3
18. In the case of transverse cracks a count of the Figure 5 Block cracking Figure 1 6 Transverse cracking Figure 1 7 Longitudinal Cracking 10 number of cracks detected in a section length was to be reported Finally if the complete data analysis and reporting could not be performed in real time then at least a reasonable 1 mile length of data should be recorded in real time It could later be downloaded and further analysis and reporting performed 10 CHAPTER TI FEASIBILITY 2 1 Sampling and Update Rates The first question to be addressed in phase one was whether or not the lasers could provide measurements at a sufficiently fast rate That is did the laser update rate meet or exceed the necessary sampling rate Since the smallest cracks to be detected were in the 1 8 inch wide range it was reasonable that a 1 16 inch sampling rate would be required The update rate of the Selcom laser system is fixed with jumpers on the receiver averaging board in the CPU sub rack This is discussed in Chapter III However the maximum update rate no averaging is 32 000 samples per second 4 5 The necessary sampling rate for 1 16 inch sampling varies from 2816 samples per second at 10 miles per hour to 14080 samples per second at 50 miles per hour A comparison of the update rate to the maximum required sampling rate shows that the Selcom lasers are able to supply measurements at the necessary speed Also since the updat
19. NPTSXBAR number of points to use in mean MTHRESH threshold value 6 NBASE number of lookahead points XBAR running mean Qe e e e e e e e he e e e e e de e e e Jx e Jx e e e e e e e dx e PROGRAM RMST CHARACTER 24 FILEIN FILEOUT DIMENSION XM 8 XD 8 XU 8 INTEGER U D IC C WRITE Input file for detect READ 901 FILEIN 901 FORMAT A24 WRITE Output file for detect READ 901 FILEOUT WRITE Number of data points READ NPTS WRITE Number of points to use in mean READ NPTSXBAR WRITE Mean threshold READ MTHRESH WRITE Slope base length READ NBASE OPEN UNIT 2 FILE FILEIN STATUS OLD OPEN UNIT 3 FILE FILEOUT STATUS NEW 6 Initialize XBAR and XM array 61 READ 2 X XBAR X XTOT XBAR NPTSXBAR DO 30 I 1 NPTSXBAR XM 1 X CONTINUE M 1 DX 200 WRITE 3 DX L 2 U 0 D 0 IC 0 C 0 Loop over all points DO 500 K 2 NPTS READ 2 X Check if X going up or down IF X LT XBAR THEN Going down so reset UP counters and arrays IF U NE O THEN 16 0 0 DO 100 J 1 U XTOT XTOT XM M XU J XM M XU J M M 1 IF M GT NPTSXBAR M 1 DX 200 WRITE 3 DX CONTINUE XBAR XTOT NPTSXBAR U 0 ENDIF Increment DOWN count store in array and check if surpasses th
20. R INTRODUCTION 5 9 5 9 e e 5 1 1 Project and Report Scope 1 2 Background s s e s 5 so 1 3 Project Phases ws e 1 4 Distress Types 9 e 1 5 Project Requirements 6 FEASIBILITY EA A 2 1 Sampling and Update Rates 2 2 Resolution Noise and Texture 2 3 Could Cracks Be Detected 2 4 The Real time Issue 2 5 Laser Problems and Limitations CRACK IDENTIFICATION HARDWARE 3 1 Optocator s e 9 9 0 3 2 68000 DAQ Board 3 3 COMPAQ Portable III vii I II III 28 28 29 30 31 32 33 36 36 36 37 38 41 46 49 52 59 65 IV TIME AND FREQUENCY ANALYSIS TECHNIQUES 4 1 Dime Series ere da 4 2 Stochastic Process 4 3 Ergodicity and Stationarity 4 4 Statistical Estimates 4 5 Power Spectrum Estimation 4 6 Linear Parametric Modeling V ANALYSIS OF PAVEMENT CRACKING DATA Sed Introduction s ea Y e w a e3 5 2 Variance Method for Real Time Crack Counting 4 v 5 s 6 5 3 Spectral Analysis Results 5 4 Running Mean Slope Threshold Method 5 5 Autocorrelation Difference Method 5 6 AR Process Modeling Results VI CONCLUSIONS AND FURTHER RESEARCH APPENDIX A DAQ BOARD SCHEMATICS 5 5 5 9 5 9 5 9 5 5 RUNNING MEAN SLOPE THRESHOLD LISTING REFERENCES 5 5 5 e 5 9
21. T XM M XU 1 XBAR XTOT NPTSXBAR XM M XU 1 M M 1 IF M GT NPTSXBAR M 1 DX 200 WRITE 3 DX DO 230 J 1 NBASE 1 XU J XU J 1 CONTINUE U NBASE 1 ENDIF ENDIF ENDIF CONTINUE Through all data points account for any 200 nana 210 220 230 500 64 data points left in UP or DOWN arrays IF U GT 0 THEN DO 600 I 1 U DX 200 WRITE 3 DX CONTINUE ENDIF IF D GT 0 THEN DO 700 I 1 D DX 200 WRITE 3 DX CONTINUE ENDIF STOP END 600 700 REFERENCES Walker Roger S Freddy L Roberts and W Ronald Hudson A Profile Measuring Recording and Processing System Research Report 73 2 Center for Highway Research The University of Texas at Austin April 1970 Claros German J W Ronald Hudson and Clyde E Lee Performance of the Analog and the Digital Profilometer with Wheels and with Non Contact Transducers Research Report 251 3F Center for Highway Research Bureau of Engineering Research The University of Texas at Austin April 1985 Walker Roger S and John Stephen Schuchman Upgrade of 690D Surface Dynamics Profilometer for Non Contact Measurements Research Report 494 1F The University of Texas at Arlington December 1986 Selcom Technical Manual Selective Electronic Co unpublished Selcom User s Manual Selective Electronic Co unpublished Cox Gregory M Donald Fronek and Rahn Merrill Real time Computer Vision Using Intelligent Hardware
22. Technical Report Documentation Page 1 Report No 2 Government Accession Ne 3 Recipient s Catalog No FHWA TX 89 1141 1 4 Title and Subtitle 5 Report Date December 1988 The Use of Lasers for Pavement Crack Detection i 6 Performing Organization Code B Performing Organization Report No 7 Author s 1141 1 Lynda Donnell Payne Roger 5 Walker 9 Performing Orgonization Name and Address 10 Work Unit No TRAIS The University of Texas at Arlington TT Convrect ec Grani No Arlington Texas 76019 Study 8 18 89 1141 7 13 Type of Report and Period Covered 12 Sponsoring Agency Name and Address Interim Texas State Department of Highways and Public Transportation D 10 Research TR nsorin en P 0 Box 5051 Austin Texas 78763 15 Supplementary Notes Study done in cooperation with US Dept of Transportation Federal Highway Administration LL 16 Abstract This research was initiated to investigate the capability of using lasers for crack detection in pavements If such a capability could be developed it would be used to aid in obtaining and evaluating pavement distress and cracking information for the State s P E S procedures used for maintaining and evaluating pavements i The research effort has involved three stages The first two stages were to determine the crack detection capabilities of the laser probes used on the Surface Dynamics Profilometer
23. al Analysis IEEE Press New York 1978 Cooley J W and J W Tukey An Algorithm for the Machine Calculation of Complex Fourier Series Math Comput vol 19 pp 297 301 April 1965 Welch Peter D The Use of Fast Fourier Transform for the Estimation of Power Spectra A Method Based on Time Averaging Over Short Modified Periodograms IEEE Trans Audio and Electroacoust vol AU 15 pp 70 73 June 1967 Ahmed Nasir and T Natarajan Discrete Time Signals and Systems Reston Publishing Company Inc Reston Virginia 1983 2 10 11 12 13 44 15 16 17 18 67 Digital Signal Processing Committee ed Programs for Digital Signal Processing IEEE Press Net York 1979 Koopmans L H The Spectral Analysis of Time Series Academic Press New York 1974 Graupe Daniel Time Series Analysis Identification and Adaptive Filtering Robert E Krieger Publishing Co Malabar Florida 1984 Marple S Lawrence Jr Digital Spectral Analysis With Applications Prentice Hall Inc Englewood Cliffs New Jersey 1987 Clements Alan Microprocessor Systems Design 68000 Hardware Software and Interfacing PWS Publishers Boston Massachusetts 1987 Coffron James William Using and Troubleshooting the MC68000 Reston Publishing Company Inc Reston Virginia 1983 Eggebrecht Lewis C Interfacing to the IBM Personal Computer Howard W Sams and Co Inc Indiana
24. arsenide GaAs laser which entails the risk of eye damage if the beam hits the eye directly 4 The GaAs laser in the gauge probe gives off pulsed modulated invisible infrared light as shown in Figure 3 2 Each pulse in the 16 pulse burst is 350 ns The bursts occur at a frequency of 32 KHz which accounts for the 32 KHz data rate of the serial data passed to the receiver averaging board The light from the laser beam passes through a lens which focuses the light in the center of the measurement range The spot size which strikes the ground surface is approximately 1 4 inch by 1 16 inch 18 19 Laser probe and probe processing unit Figure 3 1 20 The optocator measures the distance to an object by use of the triangulation principle as illustrated in Figure 3 3 From a light source L a concentrated light beam is directed onto the surface of the measured object 01 The light beam will strike the surface at point A and the scattered light reflection is focused through a lens to a point A on a position sensitive detector If the distance of the measured object is changed by X the laser beam will hit point B on surface 02 and focused at point B on the detector Since the relative position of the light source the lens and the detector are fixed the relation between x and X is known and distance measurements can be obtained The maximum measurement range 01 02 as well as the standoff distance must be considered when
25. at transverse cracking can be accurately identified It is believed that additional lasers must be installed to obtain data across the lane before the system will be able to provide this information Using only two lasers it is simply too likely that something in 30 to 50 feet of data will appear to be a crack even on smooth pavement With multiple lasers transverse cracking would be identified only after each laser across the lane had detected a crack within the same foot or two of data It should perhaps be pointed out that multiple lasers would also allow rutting to be detected Recall rutting is one of the seven distress types currently reported by PES Three lasers are being investigated in Phase 3 Another issue which remains unresolved is whether or not slight less than 1 8 inch cracking is accurately detected The old lasers with 3 8 by 1 8 inch spot size could not detect them The new lasers with 1 4 by 1 16 inch spot size have performed reasonably well on the laser calibration board but have not been thoroughly field tested due to the forementioned problems 49 50 Any user of this system must understand the limitations imposed by trying to detect cracking using only two narrow beams of laser light running parallel to the centerline Obviously massive amounts of information across the lane is not available Due to the nature of the sensors used cracks detected in failures and longitudinal cracking patterns will be misclassifi
26. ata Most important among these are the concept of a stochastic process a stationary process the autocovariance function of a stationary process the frequency content of a time series and linear parametric models Several classical texts are included in the bibliography and may be referenced for a more detailed treatment of the subject 11 12 13 It should be noted that all equations given in this chapter assume real valued time series Since complex valued time series are not considered the complex conjugation operator needed for the strictest definition of autocorrelation and autocovariance has been omitted 4 1 Time Series A signal which is continuous in time is a continuous time series A discrete time series is simply a sequence of measurements or observations taken at specific instants of time Often a discrete time series is a sampling of a continuous time series Typically the observations are taken at equispaced time increments and denoted x n A continuous time series may be obtained by measurements taken from a physical instrument Such a series is band limited and contains no frequencies higher than the maximum frequency response of the measuring instrument To analyze a continuous time series in discrete form the sampling interval must be determined such that all information present in the original signal is maintained This sampling rate must equal or exceed twice the highest frequency present in the signal and is general
27. can be obtained as solutions to the p 1 linear equations given by r 0 r 1 r p 1 Bal r 1 0 Yr p 1 31 0 r p r p 1 r 0 0 These linear equations are commonly referred to as the Yule Walker equations The autocorrelation matrix is both Toeplitz and Hermitian because r k r k where represents complex conjugation These properties allow more efficient solution than the standard Gaussian elimination The method for solution of the Yule Walker equations that takes advantage of these properties was developed by Levinson and is commonly referred to as the Levinson Durbin algorithm 31 32 CHAPTER 7 ANALYSIS OF PAVEMENT CRACKING DATA 5 1 Introduction The methods first investigated to identify pavement cracking are computationally intensive and cannot be performed in real time with the hardware developed in this study Each of these methods involve first filtering the data and then applying various statistical techniques to identify cracking Data is filtered to remove the low frequency content of the signal Low frequency components include such things as wheel bounce vehicle suspension effects bumps and hills in the section The two methods which consistently gave best results were the running mean slope threshold technique and the autocorrelation difference method These are discussed in detail in Sections 5 4 and 5 5 respectively Another technique considered was modeling the data a
28. ch block of data Also any difference greater than 1000 is plotted as 1000 so all information could be plotted on a reasonable scale As can be seen from the plot a threshold of 200 identifies all cracks except the one at point A on the plot Here a shortcoming in the algorithm is illustrated That is when one 16 point block ends and another begins in the middle of a small crack it may not be detected 45 1000 DATA POINTS 5 2 FT Figure 5 5 Raw data Thousands A D VALUE 46 Figure 5 6 and 5 7 taken together illustrate each of the cases described above Figure 5 6 shows the difference r 0 r 4 while Figure 5 7 shows the actual values of r 0 and r 4 For example Case III was large r 0 and small r 4 The actual values are plotted in Figure 5 7 and then Figure 5 6 can be examined to see the characteristics of the data and the actual difference value The autocorrelation difference method has been applied to several of the test sections with good results In fact cracking identified by this method compares favorably with that identified by the running mean slope threshold method One drawback of this method however is that it will not be able to accurately detect crack width 5 6 AR Process Modeling Results This technique was investigated to determine whether or not the coefficients obtained by modeling crack data as an AR process could successfully be used to classify cracking types and severity The assumption
29. d experiments a decision was made to obtain new lasers which had a reduced spot size The laser probes with the larger spot size could not detect 1 16 inch cracking and even did a poor job of detecting 1 8 inch cracking As expected the new lasers did a much better job of detecting less severe cracks Unfortunately with the new laser system came many problems and delays The new lasers showed an abnormally high sensitivity to sunlight In fact results were so bad that the laser probes and probe processing units had to be sent back for modification Following the modifications the probes were again bench tested both in the lab and outside in sunlight Results obtained indoors or in a shaded area were acceptable however once again when exposed to sunlight an abnormally high percentage of invalid data measurements were obtained Selcom technicians were again consulted This time Selcom suggested changing the F stop in the detector s lens system To determine the best F stop to use data was collected from the laser calibration board in direct sunlight Changing the F stop from its preset 1 4 position to 4 0 seemed to eliminate the invalid data problem The lasers were then field tested with mixed results Sufficient data was collected to continue the study Meanwhile the laser probes and probe processing units were once again shipped back to Selcom for further modification and calibration It should be noted that Selcom engineers have since
30. d the autocorrelation lag 0 is an estimate of the autocovariance lag 0 which by definition is the variance The autocorrelation difference method involves determining the spread between r 0 and r m calculated for every one inch 16 points block of data This difference is then compared with a threshold value As discussed previously r 0 an estimate of the variance for zero mean data is large for data with cracking r 0 will also be large if the data varies too far from the zero mean as is the case on a rough road when the filter is not able to keep the data sufficiently close to a zero mean This is illustrated in the last 100 data points plotted in Figure 5 6 r m is the autocorrelation for data points in the 16 point block which are m time lags apart m is typically 4 will decrease more rapidly if variance in the data is higher frequency that is sharp cracks Using the property r 0 gt r m and examining the four cases for relative values of r 0 and r m provides justification for this technique CASE 1 r 0 small and r m small implies a small difference and no cracking CASE II r 0 small and r m large is not possible by property r 0 gt r m CASE III r 0 large and r m small implies a large difference and cracking present CASE IV r 0 large and r m large implies a small difference and no cracking Figure 5 6 shows filtered data with the r 0 r 4 value plotted over the sixteenth point of ea
31. ded in the mean The accuracy of this method depends on the number of data points used in the mean the number of data points allowed in the lookahead for threshold violation and the threshold value itself After plotting and examining results from various types of cracking in the test sections it is believed that about 85 of the cracking can be identified using 4 points for the mean and lookahead value and 35 for a threshold level This technique performs better if the data is first filtered to remove the DC component and longer wavelengths A highpass Butterworth filter is typically applied to the raw data Figures 5 3 and 5 4 show the results of applying this algorithm to moderate and severe alligator cracking respectively 1000 5 2 feet filtered data points have been plotted in both figures Above the filtered data is a plot representing whether or not a crack has been seen Ground level is plotted at 200 on the Y axis and cracks at 100 The running mean slope threshold algorithm is included in Appendix B 5 5 Autocorrelation Difference Method The autocorrelation is a statistic which measures the correlation of data at different time increments apart Assuming ergodicity the autocorrelation lag m denoted r m tells if data points m time increments apart over a length of data are related The autocorrelation value will be approximately zero if the data is uncorrelated As shown by the power spectral analysis results of Secti
32. e rate is more than twice the required sampling rate it is suggested that the receiver averaging boards be jumpered for two point averaging This will provide a 16K update rate still exceeding the sampling rate required and at the same time reducing the noise in the measurements 2 2 Resolution Noise and Texture The laser measurement range as explained in Chapter III is 10 04 inches The analog signal from the laser probes varies from 0 to 10 volts A 12 bit A D converter in the probe processing unit PPU converts the analog signal into a 12 bit digital representation providing a 2 44 mV or 00245 inch resolution Noise is a major consideration in determining measurement accuracy and the ability to detect cracking That is how much variability in measurement readings would be expected if the laser was reflecting off a surface at a constant distance 11 12 To determine this the range and variance of two data sets was considered In the first the lasers were bench mounted in the lab and data was collected with the laser beam reflecting off a flat stationary object Results from this procedure showed a range of 28 9 to 32 1 mV from the mean with a standard deviation of 7 9 mV A second set of data was collected in the profilometer with the motor running and the van at rest Here the range was 36 0 to 37 2 mV from the mean and a standard deviation of 16 8 mV was observed These observations were needed to provide insigh
33. ed as alligator or block cracking There 5 little that can be done to prevent this using lasers as sensing devices The distress types such as failures patching and longitudinal cracking can only be detected if the entire lane is examined using video cameras as described by other researchers 6 7 Video system provide much more detail but this extra detail presents problems in processing out the unwanted information A system with a small cluster of lasers along and in between each wheel path would seem to provide the best choice however would likely be to costly As pointed out numerous times the algorithms developed for detailed identification and analysis cannot be performed in real time with the hardware developed in this initial study Prototype boards which are wirewrapped such as the DAQ board built for this project are limited to clock speeds less than 10 MHz because of noise problems regardless of the maximum clock frequency allowed Therefore to obtain faster speeds printed circuit boards must either be built or purchased Also to obtain more computing power a 32 bit microprocessor should be considered over the 16 bit 68000 It is believed the open architecture VMEsystem developed by Motorola should provide needed hardware upgrades for this project The VMEsystem allows the user to purchase a basic cardcage which has the VMEbus interconnect standard The user can then configure the system for his specific needs by purchasing
34. f known width and depth was needed for testing To provide this known surface the laser calibration board was built The laser calibration board though simple in concept and construction provided valuable information This board was simply a circular piece of black plywood suspended from a variable speed motor Cracks of different widths and depths were cut into the board surface The board was cut with a 2d desired circumference so it could easily simulate a road surface passing under the laser probes at speeds from 1 to 30 miles per hour by varying the rotational speed Three different sets of cracks were cut into the board Cracks within each set were the same depth That is one set of cracks was 1 8 inch deep one set was 1 4 inch and the third set was 3 8 inch in depth Five cracks of varying width were cut in each set They were 1 inch 1 2 inch 1 4 inch 1 8 inch and 1 16 inch Figure 2 1 shows the bench mounted laser probe PPU and the laser calibration board One important observation which came to light while working with the calibration board was that orientation significantly affected measurements As will be discussed in Chapter III the laser beam which strikes the measured surface is longer in one direction than the other It was found that the ability to detect slight cracking was significantly improved by having the laser beam fall across a crack perpendicular to the centerline instead of into the crack That is or
35. h sampling requires a processing time less than 71 microseconds Laser probe and laser calibration board 1 2 igure F 14 Orientation 1 Orientation 2 Direction of travel Direction of travel Crack Laser Spot Laser Spot Figure 2 2 Laser orientations 16 2 4 A D VALUE Thousands 2 3 2 2 1150 DATA SAMPLES Figure 2 3 Calibration board results Two revolutions of the calibration board are represented in the plot above Note 3 sets of cracks with 5 cracks each are included in each revolution Details of depth and width are described on page 17 a int The system described in this study has the ability to give an approximate crack count in real time or to collect a section of data in real time which will later be downloaded analyzed and reported off line The real time crack count feature is based on a variance calculation of one inch increments of data These calculations can be performed in approximately 40 microseconds It should be emphasized that this is only an estimate of cracking and is very sensitive t variance threshold values supplied by the operator Chapter VI will address the real time issue again ina discussion of upgrades and further research 2 5 Laser Problems and Limitations Initial work in determining the sensitivity of the lasers to pavement cracking used the Selcom lasers installed in the profilometer Based on results obtained from the calibration boar
36. hway Administration December 1988 The contents of this report reflect the views of the authors who are responsible for the facts and accuracy of the data presented herein The contents do not necessarily reflect the official views or policies of the Federal Highway Administration This report does not constitute a standard specification or regulation There was no invention or discovery conceived or first actually reduced in the course of or under this contract including any art method process machine manufacture design or composition of matter or any new and useful improvement thereof or any variety of plant which is or may be patentable under the patent laws of the United States of America or any foreign country 11 PREFACE This project report presents interim results from Project 8 18 87 1141 The Project was initiated to determine the feasibility of using lasers for developing an automated pavement crack detection and identification system This report provides results of the first two phases of the research effort Special recognition is due Mr Robert Harris of D 18 for his support in initiating the project and his many contributions to this research efforts Lynda Donnell Payne Roger S Walker December 1988 iii ABSTRACT This research was initiated to investigate the capability of using lasers for crack detection in pavements If such a capability could be developed it would be used to aid in obtain
37. ientation 2 in Figure 2 2 gave much better results Also orientation 1 gave invalid data readings on the back side of pratically every crack Invalid data is typically caused by an insufficient amount of laser light falling on the detector Orientation 2 showed no invalid data This observation can be explained by the fact that the entire beam fell into the crack in orientation 1 and the path of the reflected light back to the detector was obstructed by the crack wall as the beam neared the back side of the crack Figure 2 3 provides a plot of laser measurements obtained from the calibration board at 15 miles per hour using orientation 2 It can be seen that the 1 inch down to the 1 8 inch cracks are easily recognized However the 1 16 inch crack is not as easily detected In fact its true depth is not reflected in the plot The reason is that the distance value represents the average distance measurement of all tne area covered by the laser spot Since the beam does not completely fall into the crack the true depth of slight cracking cannot be accurately measured This will cause a problem because slight cracking can easily be lost in the variability seen in noise and texture 2 4 The Real time Issue The ability to detect and provide detailed analysis of pavement cracking at highway speeds up to 50 miles per hour cannot be performed by the hardware built in this initial study Real time analysis at speeds of 50 miles per hour with 1 16 inc
38. individual VMEmodules which simply plug into the VMEbus with the widely accepted eurocard connector Typical VMEmodules are microprocessor boards memory boards various controller boards and I O boards The VMEsystem architecture allows the user to configure a multiprocessor system with both local and shared memory A multiprocessor VMEsystem is currently being assembled for this project For this system VMEmodules with the 68020 microprocessor interface to the PC Each of these VMEmodules will be dedicated to processing the data from a single laser This system should provide the computing power needed to filter the data and identify cracking at least with the autocorrelation difference method in real time It is still questionable whether or not the running mean slope threshold method which provides severity information will run in real time It may very well be the case that the data will be filtered and cracks identified in 51 real time but severity information obtained off line from a reduced data set stored on the COMPAQ a reduced data set is required data compression techniques will need to be investigated further Several other methods are yet to be investigated which may aid in identifying cracks Once specific algorithms have been identified the generality of the 68020 microprocessors may not be required and a system using special purpose signal processing chips may be possible APPENDIX A DAQ BOARD SCHEMATI
39. ing Moderate transverse cracking was not included because by using 1000 data points 5 2 feet a single crack may or may not have been seen in the data thus it would appear as block or no cracking A typical power spectrum for these three types is shown in Figure 5 1 Three important observations can be made from that figure First no cracking appears as virtually a straight line There are no frequencies or range of frequencies which are predominant A flat power spectrum indicates white noise that is the signal is completely random and there is no correlation in the data A second observation is that data with cracking shows no noticeable peaks at any frequencies but does consistently show more power at the lower frequencies greater than 1 4 inch wavelengths This suggests that data 38 with cracking is correlated and statistical measures such as autocorrelation and autocovariance will be appropriate Finally moderate alligator shows more power than moderate block at the low frequencies This implies that more cracking means more correlation of the data and larger autocorrelation and autocovariance values should be observed Figure 5 2 shows typical power spectral results for sections with moderate versus severe alligator cracking Initially it was felt that perhaps different severity widths of cracking might show peaks at different frequencies This has not been observed However consistent with previous results a higher degree
40. ing and evaluating pavement distress and cracking information for the State s P E S procedures used for maintaining and evaluating pavements The research effort has involved three stages The first two stages were to determine the crack detection capabilities of the laser probes used on the Surface Dynamics Profilometer SDP The SDP is owned by the State and used for road profile measurements After experiments indicated that these probes could be used for such detection a system was developed to further study this capability and to determine how it could be used to implement an automated high speed crack identification system The third stage is the implementation of such a system so it s usefulness for P E S data collection activities can be determined This research report describes the first two phases of the research effort KEY WORDS Surface Dynamics Profilometer SDP Lasers Pavement Distress Measurements Pavement Crack Identification and Recording iv SUMMARY This project was initiated to determine the feasibility of using the laser probes on the Surface Dynamics Profilometer SDP owned by the the State Department of Highways and Public Transportation SDHPT for crack detection and identification If found feasible a system was then to be developed for use on the ARAN measurement vehicle also owned by the State so it could be used to aid in pavement distress measurements The SDP was selected for the initial testing
41. is a surface depression in the wheel paths Rutting stems from a permanent deformation in any of the pavement layers or subgrade It is usually caused by consolidation or lateral movement of the materials due to traffic loads Refer to Figure 1 1 Patches shown in Figure 1 2 are repairs made to pavement distress The presence of patching indicates prior maintenance activity and is thus used as a general measure of maintenance cost A failure is a localized section of pavement where the surface has been severely eroded badly cracked or depressed Failures are important because they identify specific structural deficiencies which may pose safety hazards See Figure 1 3 Alligator cracking is a series of interconnecting cracks caused by fatigue failure of the asphalt surface under repeated traffic loading The cracking initiates at the bottom of the asphalt surface where tensile stress and strain is highest under a wheel load The cracks propagate to the surface initially as one or more longitudinal parallel cracks After repeated traffic loading the cracks connect forming polygon shaped sharp angled pieces that develop a pattern resembling chicken wire or the skin of an alligator The pieces are usually less than 1 foot on the longest side Alligator cracking occurs only in areas that are subjected to repeated traffic loading Refer to Figure 1 4 Block cracking divides the asphalt surface into approximately rectangular pieces The bloc
42. ks range in size from approximately 1 foot square to 100 feet square See Figure 1 5 Cracking into larger blocks are generally rated as longitudinal and transverse cracking Block cracking is caused mainly by shrinkage of the asphalt concrete and daily temperature cycling It is not load associated although load can increase the severity of individual cracks This type of Figure 1 1 Rutting Figure 1 2 Patching ing tor crack iga All 4 Figure 1 Failure 3 1 igure F distress differs from alligator cracking in that alligator cracks form smaller many sided pieces with sharp angles Also unlike block cracks alligator cracks are caused by repeated traffic loadings Transverse cracking seen in Figure 1 6 consists of cracks or breaks which travel at right angles to the pavement centerline Transverse cracks are usually caused by differential movement beneath the pavement surface They may also be caused by surface shrinkage due to extreme temperature variations Although transverse cracks may occur at any spacing they will be only considered such for this research if they occur at distances greater than 10 feet apart More closely spaced cracks are counted as either alligator or block PES data and SDHPT experience suggests that this assumption will cause only a minor error in statewide PES sections Longit inal cracks are parallel to the pavement s centerline cr laydown direction They may be cau
43. ly referred to as the Nyquist rate 14 A signal from which the series was obtained could be deterministic or stochastic in nature it is possible to predict future values of the series exactly the signal is deterministic If future values can only be approximated based on statistical characteristics of past observations the signal is a statistical or stochastic time series 28 29 4 2 Stochastic Process The possible values of the time series at a given time t are assumed to be described by a random variable X t and its associated probability distribution An observed value x t at time t represents one of the infinite number of possible values of the random variable X t The probability distribution function F x t defined by F x t Prob X t gt x t is the probability that random variable x t has a value less than or equal to x t The behavior of the time series at all sampling times is described by an ordered set of random variables X t The statistical properties of the time series are described by associating a probability distribution function with each random variable in the set The ordered set of random variables X t and the associated probability distribution functions is called a stochastic process An observed time series x t is only one of an infinite number of possible realizations of the stochastic process The collection of all sequences that could result as realizations of the stochastic proce
44. mounting the laser probes Selcom s gauge probe type 2008 requires a standoff distance of 355mm 13 98 inches and has a measurement range of 256mm 10 08 inches 5 Therefore to obtain correct measurements the laser probes should be mounted such that the distance from the bottom of the probe to the ground surface middle of the measurement range is approximately 14 inches When correctly mounted distances plus or minus 128mm 5 04 inches from the calibrated ground level can be accurately measured Refer to Figure 3 4 Measured surfaces which do not fall within the measurement range will result in invalid readings The PPU processes the analog signal from the laser probe It applies bandpass and anti aliasing filters to the signal The PPU converts the analog signal into a serial digital form which can be transmitted over long distances to the receiver averaging boards located in the CPU sub rack The serial digital output includes the 12 bit value from the analog to digital converter as well as 3 invalid data bits The probe processing unit determines invalid data if the reflected laser beam is not correctly detected by the position sensitive detector in the probe For example if the measured surface is out of the measurement range the invalid data bits would reflect this and the data could be processed accordingly Another function of the PPU is to control the intensity of the laser light emitted by the GaAs laser diode in the probe
45. on 5 3 data with cracking is correlated Data with sharp cracks will show large correlation for a lag or two but the autocorrelation value decreases rapidly as the number of 1365 increases Data with longer wavelength components such as bumps show high autocorrelation values for longer lag times Section 5 2 discussed a quick and dirty way of identifying cracks in unfiltered data by calculating the variance c 0 every 16 data points and then comparing that value to a threshold That method was at best an estimate However because the data was not filtered and only a simple variance calculation was needed it did meet the real time 42 TY WT T Dar 1 AL W 0 V pda 7 1000 DATA POINTS 5 2 FT Figure 5 3 Running mean slope threshold technique applied to moderate alligator cracking data 43 1000 DATA POINTS 5 2 FT Running mean slope threshold technique applied to severe alligator cracking data Figure 5 4 44 requirement The autocorrelation difference method is an enhancement of the simple variance method Using this method the data is first filtered with a highpass filter Filtering removes the DC component and much of the variability caused by hills tire bounce and vehicle suspension effects This can be seen by comparing the raw data plot in Figure 5 5 with the plot of filtered data in Figure 5 6 With the DC component removed the data now approximates a zero mean process an
46. ould be performed in real time with the van moving at highway speeds real time computation was not feasible what procedures could be developed to collect data for later processing Phase two was to begin once it was determined that cracks could be detected using the laser probes This phase would involve designing testing and implementing both the hardware and software for a system which could be used for crack detection and identification Although it has been stated that phase one was first investigated followed by phase two this was not exactly the case Obviously some of the issues in phase one could only be addressed if there existed hardware and software to obtain the cracking data In actuality the phases overlapped and some of the hardware and software developed will be changed later based on results obtained By the same argument the success of such systems can only be determined by actual implementation 1 4 Distress Types This research is only concerned with distress types in asphalt surfaced pavements since this type of road surface represents the largest percentage of the highway system in Texas i The distress types which are currently recorded by PES raters on asphalt pavements are rutting patching failures alligator cracking block cracking transverse cracking and longitudinal cracking 8 Each type will be described here for completeness however not all types are considered in this research A rut
47. out the state to rate pavement surfaces by walking the roads Obviously this process is very tedious and time consuming Also since so many people are involved in the evaluation the ratings are often not repeatable An automated measurement system is needed to simplify the process and to obtain more consistent measurements This research represents the first attempt by the SDHPT to automate the process The research was made possible when laser probes were purchased for use on the Surface Dynamics Profilometer SDP The SDP is used by the Department to obtain road profile measurements Two lasers one in each wheel path are used to measure distances from the bottom of a survey vehicle to the road surface These distance measurements along with vertical acceleration measurements from two accelerometers are used to obtain the road profile by removing the effects of the vehicle suspension system 1 2 3 The laser system discussed in this study is built by Selective Electronic Co Selcom of Sweden The device is called an optocator The system s basic components are the laser probes and probe processing units which are mounted under the van and the CPU sub rack containing the power supply and receiver averaging boards which are installed inside the van The optocator measures distances to a surface using laser probes Each probe emits a small infrared light beam that strikes the surface to be measured The reflected light is focused
48. polis Indiana 1983 MC68000 16 32 Bit Microprocessor Motorola Semiconductors Austin Texas 1985 MC68230 Parallel Interface Timer PI T Motorola Semiconductors Austin Texas 1983 The TTL Data Book Vol 2 Texas Instruments Inc Dallas Texas 1985 Uffenbeck John The 8086 8088 Family Design Programming and Interfacing Prentice Hall Inc Englewood Cliffs New Jersey 1987 19 20 21 22 23 24 25 26 27 25 29 68 Wilcox Alan D 68000 Microcomputer Systems Design and Troubleshooting Prentice Hall Inc Englewood Cliffs New Jersey 1987 Durbin J The Fitting of Time Series Models Rev Inst Int de Stat vol 28 pp 233 244 1960 Levinson N The Weiner Root Mean Square Error Criterion in Filter Design and Prediction J Math Phys vol 25 pp 261 278 1947 30 31 32
49. put driving white noise series w n and the observed output time series x n are related by the linear difference equation 34 x n bgw n byw n 1 baw n q a x n 1 apX n p This may be rewritten in the form Lo q x n E ajx n i E b w n i i 1 i 0 This general regression model is called an autoregressive moving average ARMA model If all 84 0 then biw n i 0 x n Ma i and the process is known as a moving average model of order q and represented MA q If all bj 0 i gt 0 then i x n 4 8 3 bow n 1 and the process is known as an autoregressive model of order p that is AR p Any one of the three parametric models described above may be expressed in terms of the other two models An ARMA or MA model of a finite number of parameters may be described by an AR process generally of infinite order Similarly an ARMA or AR process can be expressed as a MA model of infinite order This observation is important because it suggests that any of the three models may be selected and a reasonable model obtained if a sufficiently large order is used Of the three models the AR model has mathematical characteristics which have allowed the development of a number of efficient algorithms Specifically AR models have linear solutions whereas solving for ARMA or MA parameters involves nonlinear equations 55 Estimates of the AR parameters 84
50. reshold D D 1 XD 5 X IF XBAR X GE MTHRESH THEN DO 110 J 1 D DX 100 WRITE 3 DX 30 QOO ana QOO 100 nana 62 110 CONTINUE Check if have been in crack too long and RESET if gt 6 Q Q Q Q IF C GT 96 THEN C 0 XBAR X XTOT XBAR NPTSXBAR DO 120 J 1 NPTSXBAR XM J X 120 CONTINUE M 1 ENDIF D 0 ELSE If were decreasing for NBASE number of of lookahead points but did not surpass threshold then update by 1 point and cont Q Q Q Q 0 IF D EQ NBASE THEN 0 0 0 XTOT XTOT XM M XD 1 XBAR XTOT NPTSXBAR XM M XD 1 M M 1 IF M GT NPTSXBAR M 1 DX 200 WRITE 3 DX DO 130 J 1 NBASE 1 XD J XD J 1 130 CONTINUE D NBASE 1 ENDIF ENDIF Ckkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkk 0 6 Similar code as for DOWN but here are going 6 UP 0 O e e de e he e de de e dee Xi XXX Xi XX XXX ELSE IF D NE O THEN IC 0 0 DO 200 J 1 D XTOT XTOT XM M XD J XM M XD J M M 1 IF M GT NPTSXBAR M 1 naaa 63 DX 200 WRITE 3 DX CONTINUE XBAR XTOT NPTSXBAR D 0 ENDIF U U 1 U X Make X surpass 3 threshold before kick out as data point not to be included in running mean IF X XBAR GE 3 MTHRESH THEN DO 210 J 1 U DX 300 WRITE 3 DX IC IC 1 CONTINUE Checking if need to reset IF IC GT 96 THEN IC 0 XBAR X XTOT XBAR NPTSXBAR DO 220 J 1 NPTSXBAR XM J X CONTINUE M 1 ENDIF U 0 ELSE IF U EQ NBASE THEN IC 0 C 0 XTOT XTO
51. s an AR process and then examining the AR coefficients This method would allow crack identification and classification if each cracking type would give distinctly different AR parameter values and the same type cracking would give similar coefficients The AR modeling results are discussed in Section 5 6 As stated the methods mentioned above give detailed analysis and cannot be performed in real time with existing hardware It was desired to develop a technique even a rough estimate which could perform in real time with the hardware described herein A technique using a variance measure has been implemented which provides a crack count in real time This is discussed in the following section 5 2 Variance Method for Real Time Crack Counting Although detailed crack identification and classification cannot be obtained using the DAQ board and COMPAQ at highway speeds an estimate of the number of cracks seen is possible using a simple variance calculation This method simply calculates the variance every 16 data points 1 inch and compares that statistic to a threshold level provided by the operator If the variance for that inch of data surpasses the 36 37 threshold the count is incremented and displayed on the COMPAQ This calculation takes approximately 40 nicroseconds on the DAQ board well within the 71 microsecond requirement for 50 miles per hour The variance is calculated on 16 raw data values Since unfiltered data is
52. scribes the work done to determine if pavement cracking could be detected with the lasers The third chapter describes the hardware designed and built for initial evaluation of a crack detection system Chapter four defines and explains the statistical and signal processing theory used in the crack identification algorithns Chapter five describes the different crack identification algorithms employed This chapter also describes results of the data analysis on the test sections used in the study Chapter six describes additional research much of which is being conducted in the third phase 1 2 Background The evaluation of pavement surface conditions of the nation s highways is of major interest to transportation engineers The State has been using such information in conjunction with other data in an established procedure for determining the condition of the State s highway system This information is essential in determining which roads should be worked on and how much money is needed to complete the work The State currently evaluates pavement surface conditions by considering both road roughness and pavement distress A measure of road roughness is readily obtained with existing instruments Pavement distress information is more difficult to obtain as it requires visual evaluation Currently SDHPT personnel attend the annual Pavement Evaluation System PES Rater Training School and then disperse to their respective districts through
53. sed by Welch 17 This method segments the data applies a window to each segment determines the periodogram of each windowed seqment and then calculates the average periodogram which is called the modified periodogram With this method the data segments may be overlapped This method of periodogram averaging reduces the variance of the spectral estimate The essential features of this method are described below The available time series x n 0 lt n gt N 1 is divided into K overlapping segments of length L The segments overlap by L 2 samples The total number of segments then becomes K N L 2 L 2 where any fractional portion of K is truncated The ith data segment then becomes 33 xj 8 x iL 2 n w n where 0 gt n gt L 1 0 gt i gt 1 1 and w n is a window function of length L Typically either a rectangular or Hamming window is used The DFTs of each of the K data segments are then computed using the FFT algorithm by x k E x n exp ikn 2M M where 0 lt k lt M 1 and 0 lt i lt K 1 M is the DFT length and must be gt L The modified periodograms S k are then averaged to produce the spectrum estimate S 2Wk M 1 KU E Si k for 0 lt k lt M 1 0 lt i lt K 1 and Si k X4 X and L 1 U w n n 0 i 4 6 Linear Parametric Modeling Many discrete time stochastic processes can be approximated by a linear regression model In this model the in
54. sed by a poorly constructed paving lane joint shrinkage of the surface due to low temperatures or hardening of the asphalt or a problem with the subgrade Refer to Figure 1 7 Note the figure also has block cracking This research effort considered only three of the seven distress types described above Specifically alligator block and transverse cracking were to be considered Some of the other distress types particularly failures and longitu inal cracking could cause the cracking pattern to be misclassified due to the nature of the sensors used and the method of observation This should become clear from later discussions 1 5 Project Requirements As previously described this study involved using the existing lasers to identify cracking patterns One laser was to be mounted in each wheel path and one in the middle Obviously little or no information across the lane could be recorded to help in the identification The laser data was to be recorded and analyzed in real time at highway speeds if possible The type severity and percent area of cracking was to be determined from the laser data obtained Type refers to one of the three types previously mentioned alligator block or transverse Severity is determined by the width of the crack Slight cracks are less than 1 8 inch moderate are 1 8 to 1 4 inch and severe are greater than 1 4 inch wide Also the percent of the section with each type of crack was to be noted
55. ss is called an ensemble of sample sequences The expectation of a random variable X t at time t denoted by E X is given by E X x p x dx x 6 Here x is the observation at time t and p x is the probability density function of X t This implies that the mean X is based on values x taken from all possible ensembles of the random variable at time t The expected value of the squared magnitude of random variable X is E x x p x dx gt is the mean squared value of X The variance of a random variable is the mean squared deviation of the random variable from its mean 30 var X x E x p x dx 8 8 E X Ex An indication of the statistical relationship of one random variable X1 at time t1 to another X2 at time t2 is given by the autocorrelation r X1X2 E X1X2 This represents the engineering definition for autocorrelation as first suggested by Weiner The autocorrelation of a stochastic process with the mean removed is the autocovariance given by E X1 E X1 X2 E X2 c X1X2 r X1X2 X1 X2 If the random process has zero mean for time t1 and t2 then c X1X2 r X1X2 Also if the random variables X1 and X2 are mutually independent or uncorrelated then C X1X2 o This implies that there is no relationship between the two random variables and knowing values for X1 does not help in predicting a value of X2 4 3 Ergodicity
56. t into reasonable threshold values used in several of the crack detection algorithms The texture of a road surface is another item which adds variability In fact it should be understood that road surfaces of very course texture probably do not allow reasonable crack detection by the methods described in this study 2 3 Could Cracks Be Detected Phase one of this project involved determining whether or not the Selcom lasers on the profilometer could detect cracks in a road surface Two approaches were taken to answer this question First short sections of pavement with the desired cracking were located The sections were marked as to start end and the desired path for the driver to take Laser data was then obtained from the sections with the driver being very careful to follow the marked path This data was plotted and compared with slides taken of the marked section Results of this comparison were very encouraging Most of the moderate and severe cracks seen in the slides could easily be recognized in the plots The first procedure of driving over a marked section gave a good idea but it was never known exactly where the laser beam fell That is a crack perpendicular to the centerline may be 1 4 inch wide at one point while 1 2 inch over it might be 1 16 of an inch wide For this reason that procedure did not give much insight into how well the lasers would be able to provide severity information Therefore a surface with cracks o
57. the next point If it is less than a threshold distance below the running mean then it is not a crack and the value replaces the oldest value used in the mean calculation and a new running mean is determined Unfortunately this will not provide accurate crack identification for cracks with gently sloping walls The problem is that although the values are decreasing they may not exceed the threshold using the technique described above and so they are included in the running mean This lowers the mean value and makes it even more difficult for the next point to be identified as part of a crack The 39 100 90 80 70 60 50 40 30 20 10 0 0 062 0 125 0 187 0 25 0 312 0 375 0 437 05 FRACTION OF SAMPLING FREQUENCY No Cracking Mod Block o Mod Alligatar PSD db a Figure 5 1 Power spectral density plots of different cracking types 100 90 PSD db 0 0 062 0 125 0 187 0 25 0 312 0 375 0 437 0 5 FRACTION DF SAMPLING FREQUENCY a Mod Alligator Sev Alligator Figure 5 2 Power spectral density plots of different cracking severity 41 problem is solved by looking ahead up to L lookahead points assuming all values are constantly below the mean for a value exceeding the threshold before updating the running mean L is a user supplied parameter If the threshold is exceeded within L points then each of the decreasing data points are identified as part of a crack and will not be inclu
58. used there exists variance in the data due to the factors previously mentioned which contribute to low frequency content of the signal However because only 16 data points are used in each calculation these components do not contribute as heavily to the variance value as high frequency cracks and thus filtering can be neglected to save calculation time The accuracy of this technique is highly dependent upon the operator entering meaningful threshold values More investigation is needed to determine reasonable threshold limits for various pavement textures 5 3 Spectral Analysis Results Typically one of the first things that should be considered about any measured signal is its frequency content As previously discussed spectral analysis provides this information Of particular interest in this study was a determination of whether or not the different cracking types displayed characteristic power spectra Also it seemed reasonable that cracking of the different severity types might show characteristic peaks at different frequencies The procedures described below provide information about the frequency content of pavement cracking data The first question addressed was whether or not each cracking type had its own characteristic spectrum Here several data segments of 1000 data points in length were identified from the test sections for each of the desired types The types considered were moderate alligator moderate block and no crack
Download Pdf Manuals
Related Search