Automated Pavement Distress Collection Techniques (2004)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

8This chapter is devoted to a discussion of various pavement distress data collection issues and the technologies involved. To organize the voluminous data from the questionnaire responses and the literature search into a manageable format, several issues, common to all distresses, are discussed early in the chapter. These are monitoring frequency, sampling or reporting frequency, and location-reference methods. Later in the chapter, the various automated data collection technologies and equipment currently in use by highway agencies are identified and discussed. The equipment used generally conforms to the provisions of the ASTM Standard Guide for Classification of Automated Pavement Condition Survey Equipment (16). As suggested earlier, the data collec- tion technologies fall into two general classes: imaging of the pavement surface through photographing, videotaping, or digitizing: and the measurement of pavement longitudinal and transverse profile through the use of various noncontact sen- sors. The existing technologies are discussed in the order they were addressed in the project questionnaire. In addition, sev- eral emerging technologies and how they are seen to apply in highway agencies are described. Table 1 provides an over- view of pavement data collection and processing. DATA MONITORING FREQUENCY Although there are numerous variations, there tend to be dif- ferences within many agencies between the monitoring fre- quency used for pavement surface distress (imaging) and that used for the sensor-measured features (roughness, rut depth, and joint faulting). Essentially, that difference pertains to the relative difficulty in collecting and processing imaging data. Many agencies collect sensor data more frequently than images. Table 2 provides a summary of frequencies from the data gathered through the analysis of questionnaire responses and is detailed in Tables B1 through B4 in Appendix B. Table 2 reflects only those pavement data collected through automated means, whereas the tables in Appendix B cover some elements collected manually as well. Almost all agencies monitor pavement surface distress (cracking, etc.) at 1-, 2-, or 3-year frequencies, if at all. Vari- ations are that a few agencies do one-half of the system each year, with others doing one-third each year. In the tabulation, these results are expressed as 2 and 3 years, respectively. A few states monitor Interstate pavements at 1-year intervals and other pavements at 2-year intervals (Table B1). Two Canadian provinces use the 3-year interval, whereas British Columbia monitors at a 2- to 4-year frequency depending on the class of highway. Most agencies (18) capture pavement cracking at 2-year intervals, as shown in Table 2. The 1-, 2-, and 3-year monitoring frequencies apply to all other distress features captured. Almost every agency moni- tors roughness (IRI). Because that monitoring is at least partly driven by HPMS requirements, it is necessary to do the work on at least a biennial basis. Table 2 shows that the agencies are almost equally split between a 1- and 2-year roughness moni- toring frequency. Again, a few agencies monitor their Inter- state pavements at 1-year intervals and other roads at 2-year intervals. Finally, two Canadian provinces and Massachusetts conduct roughness monitoring at 3-year intervals. Again, British Columbia monitors roughness at a 2- to 4-year fre- quency, depending on the class of highway. Rut depths typically are concurrently determined with mea- surements of roughness because the same sensor technology can be used. Therefore, there is the same virtually equal split between 1- and 2-year monitoring frequencies. Four agencies reported a 3-year rut-depth monitoring frequency, whereas British Columbia again uses 2 to 4 years. Two agencies that measure roughness do not collect rut-depth measurements. Far fewer agencies employ automated collection of joint faulting than the other sensor collected parameters. In addition, because some agencies, especially the northeastern states and most of the Canadian provinces, have few exposed concrete pavements, there seems to be some lack of confidence in the automated means (as discussed later) of making the measure- ments. Several agencies collect joint-faulting data with one or more version of a manually operated faultmeter and are not included in Table 2. Of those collecting automated faulting data, 10 use a 1-year monitoring frequency and 13 use a 2-year cycle. DATA REPORTING INTERVAL An initial effort was made to address pavement condition data sampling interval (measured in longitudinal distance or per- centage of length). However, it was determined that most agencies using automated means of data collection sample continuously, or very nearly so, on the outer traffic lane. In a few instances, a worst lane is selected for evaluation. In no CHAPTER TWO DATA COLLECTION ISSUES AND TECHNOLOGIES

9case is an agency evaluating all lanes. The essentially univer- sal practice is to evaluate the outermost traffic lane (no park- ing spaces) in one direction for pavements having fewer than four lanes and in both directions for roadways having four or more lanes. For purposes of this discussion, 100% sampling means 100% of the evaluation lane. In cases where shoulders and other peripheral features are evaluated, those evaluations are conducted separately as described later. Images usually provide continuous coverage at 3 to 5 m (10 to 15 ft) longitudinally per image, whereas sensor mea- surements often are made at intervals of 25 to 100 mm (1 to 4 in.). Thus, it was deemed more meaningful to address a reporting interval, because that is a more standardized quan- tity than distance. The results of questionnaire responses con- cerning reporting intervals are given in Table 3 (details are cited in Tables B1–B4). Only those agencies reporting auto- mated collection are included in Table 3. Even then, a few agencies declined to state a frequency for at least some data items. For cracking, nine agencies reported only that they sam- pled 100% of the lane to be evaluated. Many others reported 100% sampling of that lane but listed reporting intervals of 15 to 300 m (50 to 1,000 ft). Such a statistic simply means that all data (100% of the evaluated lane) are used, but that the results are summarized at an agency-dependent frequency. Three agencies collect cracking data on a segment-by-segment basis (usually defined as a pavement management segment of varying length), whereas five sample 10% to 30% of the roadway, usually on a random sampling basis. A few agen- cies define sampling or reporting intervals in other ways, such as the 500 m employed by New Brunswick. The LTPP program is included as an agency using 100% sampling, although it should be kept in mind that each LTPP site is typ- ically 500 ft long. In the case of roughness monitoring, many U.S. agencies employ 100% sampling with reporting intervals at 158 m (0.10 mi or 528 ft). This reporting interval seems to be an “English” version of the 0.1-km interval suggested in the AASHTO IRI provisional standard (12) and covers most of the 100 to 300 m (300 to 1,000 ft) range given in Table 3. The Canadian provinces reported roughness results at 50- to 100-m intervals. A few agencies report roughness results by pavement management segment, whereas a few others use reporting intervals ranging from the one city block used by the District of Columbia to 1 mi used by Arizona. For given agencies, rut-depth measurements tend to be reported in much the same intervals as roughness. Also as cited in Table 3, there are 12 agencies simply reporting 100% sam- pling, whereas others use reporting intervals of 10 to 300 m (30 to 1,000 ft). Five report results by segment average and three use other intervals such as 1 mi (Arizona) and a sample from each mile (Oregon). Finally, far fewer agencies employ automated collection of joint-faulting data. However, as given in Table 3, the inter- vals fit those given for roughness and rut depth. Note that 11 agencies sample 100% of the joints, 5 report faulting at 100 to 300 m (300 to 1,000 ft) intervals, and 7 report average faulting by pavement management segment. Frequency (years) Cracking, etc. Smoothness/ Roughness Rut Depth Joint Faulting 1 2 3 Other Total 9 18 2 1 30 26 20 4 2 52 24 20 4 2 50 10 13 0 0 23 TABLE 2 SUMMARY OF AUTOMATED MONITORING FREQUENCIES EMPLOYED (Number of Agencies) Data Item Activity Entity/Process Cracking IRI Rutting Joint Faulting Agency Contract Agency Contract Analog Digital Laser Acoustic Infrared AASHTO ASTM LTPP Other 10 20 7 7 16 17 — — — 4 — 5 21 31 23 — — — — 44 3 4 12 19 — 16 30 21 — — — — 30 15 2 6 — — 38 21 12 — — — — 23* 4 — — 10 Automated Collection Automated Processing Image Capture Sensor Data Collection Protocol Use *By sensor. Notes: LTPP = Long-Term Pavement Performance; IRI = International Roughness Index. TABLE 1 OVERVIEW OF AGENCY PAVEMENT DATA COLLECTION AND PROCESSING (Number of Agencies)

10 LINEAR REFERENCING A major input to location-referencing systems is the linear- referencing element. The method in use by a given agency has very little relationship to the types of distress collected by that agency. Almost without exception, the reference system used for one data element is used for all elements. Table 4 is a sum- mary of the linear-referencing systems in use (details are given in Tables B1–B4). Column totals would not be meaningful because a number of agencies use more than one method at a time. The column indicating smoothness and roughness proba- bly is the best true indicator of linear reference use, because it is the indicator of the most agencies using an automated method of data collection. For example, because fewer agen- cies use mile points for joint faulting than for roughness means that fewer agencies collect joint-faulting data; it says nothing about the preferred reference methods. Clearly, there is a strong preference for the use of mile posts or mile points, and the two terms seem to be used inter- changeably, although technically there is a real difference. Mile points refer to a specific location on a roadway, whereas mile posts are the physical markers for those locations. For the purposes of the rest of this synthesis, the term “mile point” will be used to identify either the mile point or mile post des- ignation unless otherwise clearly distinguished in the text. It is also not clear that there is a difference between mile points and the log mile terminology used by several agencies. The data are summarized on the basis that there is no difference between the two. Similarities in linear-reference systems in use and the HPMS linear-reference guidelines suggest that many agencies have adopted those guidelines (5). Interest in the use of a geographic information system (GIS) and even the Global Positioning System (GPS) was identified in NCHRP Synthesis 203, although no tabulation of users was provided (6). There is currently a definite trend toward the use of GPS coordinates (latitude and longitude) for location-reference purposes, although the technology has not been broadly accepted as the only method. In all except two cases, agencies reporting the use of GPS coordinates also con- tinue to use mile points. There is some indication that this dual use may be temporary and that more agencies eventually will adopt coordinates as the sole method. However, there is some recognition that a tremendous volume of archived highway data has used only mile points, so that abandonment of that approach would require absolute certainty of good correlation of the two methods. Furthermore, discussions with highway maintenance personnel strongly suggest that physical mile posts will be in use for working purposes well into the future. PAVEMENT SURFACE DISTRESS Procedures in Use Numerous procedures for asphalt pavement crack identifica- tion and collection are in use in various agencies, although four agencies reported the adoption of AASHTO Provisional Standard PP44-01, Standard Practice for Quantifying Cracks in Asphalt Pavement Surface (17). That standard defines a crack as a discontinuity in the pavement surface with mini- mum dimensions of 1 mm (0.04 in.) wide and 25 mm (1 in.) long. It further defines a low-severity crack as less than 3 mm (0.125 in.) wide. The existing imaging technology seems capable of reliably capturing the latter cracking through auto- mated means, although there is some effort to capture even finer cracks. In a data collection contract, Alabama has sug- gested that a 0.5 mm (0.02 in.) minimum would be desirable in its system. In response, the contractor stated, with regard to its latest (November 2002) high-resolution digital camera, “We are unable to detect these (1⁄2 mm) reliably and, frankly, we do not think any system existing today can do so, most cer- tainly for network level uses and at network level prices” (personal communication, letter to S. George, Alabama DOT, from Roadware Group, Inc., Nov. 18, 2002). The contractor went on to propose a minimum 2 mm (0.08 in.) crack as a level it could be confident in achieving. Method Cracking, etc. Smoothness/ Roughness Rut Depth Joint Faulting Mile Point (post) Latitude–Longitude Link-Node Log Mile Other 33 12 5 3 2 46 15 5 1 1 35 14 5 1 1 23 8 2 0 0 TABLE 4 SUMMARY OF LINEAR-REFERENCE METHODS USED IN AUTOMATED MONITORING (Number of Agencies) Interval Cracking, etc. Smoothness/ Roughness Rut Depth Joint Faulting 100% 100–300 m (300–1,000 ft) 10–50 m (30–160 ft) Segment 10%–30% Other Total 9 6 6 3 5 1 30 12 20 13 2 0 5 52 12 16 15 3 0 4 50 11 5 3 1 0 3 23 TABLE 3 SUMMARY OF REPORTING INTERVALS USED IN AUTOMATED MONITORING (Number of Agencies)

11 Other procedures include that developed for the LTPP pro- gram (18) in use by LTPP and several other agencies, and the pavement condition index (PCI) approach developed by the U.S. Army Corps of Engineers (19) and in use by Wyoming, and as one input to the procedures used by numerous others. In those cases, the standards are being adapted to automated data collection. In addition, some 20 agencies are using agency- specific protocols for crack data collection and classification, usually by manual collection methods. Many of those agen- cies expressed an interest in the AASHTO standards, but they have not moved to its adoption. For those familiar with the AASHTO procedure, most resistance to adoption appears to be related to an unwillingness or technical inability to do away with a workable agency protocol that is not totally com- patible with the AASHTO standard. Often there are large data- bases of historical data collected through an existing protocol that may be lost to an agency when adopting the AASHTO provisional standards. Fourteen agencies provided copies of their asphalt pave- ment surface distress rating procedures in response to the syn- thesis survey. Although almost all were written for manual surveys, they are now used to support automated procedures. Because agency and AASHTO procedures typically define types of cracking with reference to vehicle wheel tracks (also referred to as wheel paths), there have been efforts to put dimensions on those areas of the pavement. The definition of wheel paths and the related survey area used by AASHTO Provisional Standard PP44-01 is given in Figure 1. Note that the wheel paths and area between them are fixed regardless of lane width. Variable lane widths are accommodated in the left and right areas outside the wheel paths. Some agencies use location with respect to the wheel paths as a determinant of fatigue cracking. Those agencies will define wheel tracks as a part of their automated distress processing methodology (see chapter three). Automated data collection is being used on PCC pavements as well. Twelve agencies provided copies of their concrete pavement distress rating procedures in response to the synthe- sis survey. Again, most were developed for manual surveys, but may be used to support automated procedures at this time. Methods of Data Capture Pavement surface distress is captured by several different methods, as summarized in Table 5. Agencies doing manual surveys are included for comparison purposes only. Although approximately one-half of the agencies reported using a man- ual collection methodology, it is clear from stated plans that automated approaches will be coming into progressively more use. Also, a few agencies are in a transition period and use some manual and some automated collection methods. A few others use manual surveys on low-traffic-volume roads and automated approaches where safety is a major issue owing to high-traffic volumes. In the past, there have been some efforts to capture pave- ment cracking through the use of acoustic or laser sensors that attempted to relate cracking to abrupt variations in pavement texture (6). Such approaches seem to have gained little favor and have lost out to the imaging methods now used. The major methods of pavement imaging are generically termed “analog” and “digital.” Analog refers to the process FIGURE 1 Cross section of survey lane showing wheel paths and defined survey area between wheel paths as defined by AASHTO (17).

12 wherein images are physically imposed on film or another medium through chemical, mechanical, or magnetic changes in the surface of the medium. Digital imaging refers to the process wherein images are captured as streams of electronic bits and stored on electronic medium. The digital bits can be read electronically for processing or reproduction purposes. A third emerging method, three-dimensional (3-D) laser scanning, will be discussed later. Although much of the earlier work has been done with analog photographs or videotapes, digital imaging is fast becoming the most popular method, owing to the quality of images that can be produced, the ease of data manipulation, and the applicability to automated data reduction (to be described later). Digital images may be captured on videotape or on other media, such as computer hard drives, compact discs, or digital video discs. Whatever means of image cap- ture is used, these images are almost always “stamped” with date, time, and some means of location-reference so the image can be tied to a given location. Pavement imaging methods are described here in more detail. Much of the discussion is derived from a draft Trans- portation Research Circular, Automated Imaging Technolo- gies for Pavement Distress Survey, provided by TRB Com- mittee A2B06, Pavement Monitoring, Evaluation, and Data Storage (20). The circular presents the concept of pavement surface distress surveys using pavement images, reproduced in Figure 2. The system consists of data acquisition, data storage, and data display and processing subsystems. In addi- tion, a database system is used for archiving and retrieving the processed data. Analog Imaging The predominant use of analog imaging of pavements is in photographing (usually with 35-mm film) and videotaping. Images obtained can be of high quality, but they are not easily converted to digital format for computer storage and manipu- lation. Analog imaging has been less frequently used in recent years owing to the maturing of digital technology. The draft TRB circular (20) sums up the analog technology. The quality of televisions and videotapes, including Super VHS (S-VHS) format videotapes and 12-in. laser discs, is determined by the analog video standard set by the National Television Sys- tems Committee in the early 1950s. Although an analog video signal can be transmitted and copied through narrow band- widths, it is difficult to manipulate, copy, and distribute the sig- nal without introducing electronic noise into the original signal, which degrades image quality. It is also difficult to integrate ana- log video with other types of data, such as text and graphics, unless high-end video production equipment is available and used. The resolution of the standard analog video signal is also relatively low compared with that of digital alternatives. There- fore, today’s highway users of videotapes have largely transi- tioned into using computer-based digital technology. Still, the photographic method, popularly known as photo- logging, was used by a few agencies for many years. It prob- ably became most well known for its adoption as the method of choice for the LTPP program now managed by the FHWA and described by Gramling and Hunt (21). The photologging methodology essentially consists of photographing the pavement surface, usually with 35-mm film, and reduction of distress data through review of the film at a workstation. Photologging vans typically use a downward- facing camera and possibly one or more facing forward or in another direction, depending on user needs. Most earlier work Image Aquisition and Compression Sub-System Surface Distress Database Image Data Storage Image Display and Processing Sub-System FIGURE 2 System concept in the pavement surface distress survey (20). Agency Type Manual Analog Photographic Video Digital State Province Federal Total 17 6 1 24 0 0 1 1 13 2 0 15 15 1 1 17 TABLE 5 METHODS OF SURFACE DISTRESS CAPTURE (Number of Agencies)

13 was done by contract on a cost-per-mile basis. Much of the work was done at night using lighted cameras to overcome problems with shadows cast by survey vehicles, traffic, or roadside features that can mask pavement features critical to proper distress evaluation. In most cases, photologging was continuous over what the agency defines as a roadway section or sample of a roadway section. At least one film image ven- dor reported the capture of cracks as fine as 1-mm (0.04-in.) wide at speeds up to 96 km/h (60 mph) when controlled illu- mination is used (22). NCHRP Synthesis 203 (6) noted that three agencies still used the photographic approach about 10 years ago. Now, only LTPP reports using photologging. LTPP makes full use of the technology on test sites throughout the United States and Canada. “Instead of shooting one frame at a time, the film used in LTPP is continuously exposed to a moving pavement surface, forming a contiguous image of pavement. In order to control illumination to guarantee image quality, shooting is normally conducted at night” (20). LTPP reported that cracks of approximately 2.5 mm (0.1 in.) or less are not consistently seen on film depending on lighting, moisture, and other con- ditions (13). Interestingly, the LTPP imaging contractor uses a unique method of estimating rut depths from a transverse line superimposed on the pavement image. That photographic method is discussed later in this chapter. The method of choice for pavement imaging by many agen- cies is a videotaping technology that consists of the capture of pavement images on high-resolution videotapes, usually of the S-VHS variety. Questionnaire responses showed that approxi- mately one-third of the agencies reporting have adopted video imaging for at least part of their pavement surface distress data capture. Typical survey vehicle configuration consists of one or more downward-facing video cameras, at least one forward- facing camera for perspective, and any number of additional cameras for the capture of right-of-way, shoulder, signage, and other information depending on agency requirements. As with photologging, pavement cameras may use special lighting to reduce shadows that can mask distress features. Reduction of distress data from videotape images also involves the use of workstations and manual review of the images to classify and quantify distresses. The method is cumbersome and has given way in recent years to digitizing of the images for more ready data handling and processing, as described in chapter three. Digital Imaging The employment of digital cameras is rapidly becoming the preferred method of pavement imaging. As with analog video- taping, slightly more than one-third of the responding agencies have begun to use digital imaging to capture pavement surface distress data. Survey vehicle configuration is similar to that for videotaping in that one or two cameras capture the pavement image while any number may be used for other data required by the agency. Again, special lighting may be used to over- come shadowing problems. A major force behind the move toward digital imaging of pavements is the opportunity to reduce distress data from those images through automated methods. Another advantage of digital imaging is the availability of random access to the data. As mentioned earlier, digital images lend themselves to automated analysis because of the ability to analyze variations in grayscale as those variations relate to pavement features. Several automated analysis methods are in use and others are under development to accomplish that task, as discussed in chapter three. There are two types of cameras currently used to digitally image a pavement surface. These are known generally as the “area scan” and the “line scan” methods, although some ven- dors are using other terminologies. Wang and Li (23) provide a good overview of digital camera pavement imaging. The National Endowment for the Humanities has funded a digital imaging tutorial developed by and available from Cornell Uni- versity (24). Furthermore, the TRB draft circular (20) includes a primer on digital imaging. The two scanning approaches are depicted in Figure 3. Area Scan This method of digital imaging refers to that in which an image consisting of thousands of pixels depicts some defined pavement area, usually one-half to full-lane width and 3 to 5 m (10 to 15 ft) long, depending on camera features (lens, camera angle, placement) and vehicle speed. In pavement imaging, camera angle is of great importance, for distorted pixels (and images) will occur if the camera is not perpen- dicular to the pavement surface. The resolution varies somewhat among agencies and ven- dors and is increasing steadily as the technology evolves. Area scanning uses a two-dimensional (2-D) array of pixels in a conventional sequence of snapshots. The three basic types of area scan arrays are full frame, frame transfer, and Inter-Line Transfer, shown in Figure 3a. Descriptions of these technolo- gies are given in the draft TRB circular (20). An example of a pavement image taken with an area scan camera is shown in Figure 4 at the resolution of 2,048 pixels transversely in Joint Photographic Experts Group (JPEG) format (20). Line Scan The most common example of line scan imaging is the fax machine. Line scan imagers use a single line of sensor pixels (effectively one-dimensional) to build up a 2-D image. The second dimension results from the motion of the object being imaged. The 2-D images are acquired line by line by succes- sive single-line scans while the object moves (perpendicu-

14 larly) past the line of pixels in the image sensor, shown in Figure 3b (20). Thus, line scan pavement imaging is performed through the digital capture of a series of transverse lines that are full- pavement-lane width. These lines are “stitched” together to form a continuous image or an image broken at intervals set by the user. The International Cybernetic Corporation (ICC) describes its line scan setup in this way: ICC’s pavement digital imaging system uses a linescan camera with a resolution of 2048 pixels by “N” lines. [N is adjustable from 130 mm (5 in.) up to 9 m (30 ft) of longitudinal pavement coverage.] The width of the image is determined by the height of the camera above the pavement. Image width is adjustable between 9 ft and 15 ft with the standard boom (25). Mandli, Inc., reports that 3-mm (0.125-in.) wide cracks are clearly visible on full-lane width line scan images (26). Wang and Li (23) provide a good discussion of the relationship between vehicle speed and the number of pixels required per line. They also note that the resolution of line scan cameras was as high as 6,000 pixels per line in 1999. At least one vendor uses the area scan approach for for- ward and right-of-way images and a line scan camera for pave- ment images at a claimed crack visibility down to 1 to 2 mm (0.04 to 0.08 in.) (22). A particularly onerous problem with line scan imaging can result from any shadows cast by the survey vehicle itself. Because of the line scan feature, any shadow from the vehi- cle that falls onto the pavement surface will appear as a con- tinuous shadow in the scanned image. If this shadow falls in a critical area of the pavement, a wheel path, for example, the image can be rendered virtually useless. Special precautions and sometimes special lighting must be used to avoid this problem with line scans. The ICC confronts the shadowing problem by using a 10-fixture lighting system for all line scanned pavement images (25). An example of a line scan pavement image is given in Fig- ure 5 and an image with a vehicle-cast shadow is shown in Figure 6. As with the other methods of imaging the pavement surface, digital images require a workstation with the appro- priate software to capture pavement cracking and other fea- 2,048 pixels (transversely), about 3 meters wide FIGURE 4 A 2,048-pixel resolution image in JPEG format (area scan digital camera) [Courtesy: Wang (20)]. (a) Area Scanning (b) Line Scanning FIGURE 3 Two scanning approaches in digital imaging [Courtesy: DALSA Corporation and Wang (20)].

15 tures for pavement management purposes. These issues are discussed in chapter three. Three-Dimensional Laser Imaging The TRB draft circular provides the following description of an evolving technology using 3-D imaging (20): Phoenix Scientific Inc. (http://phnx-sci.com) has developed a phase-measurement Laser Radar (LADAR) to measure the 3-D properties of pavements. The Laser Radar uses scanning laser and reflector to measure the reflecting times across pavement surface, therefore establishing a 3-D pavement surface after the Laser Radar moves longitudinally along the traveling direction. Its system, as claimed, is able to produce roughness and rut- ting data. Another company, GIE Technologies Inc. in Canada (http://www.gietech.com/), has the LaserVISION system, which also models the 3-D surface of pavements. The lasers are sta- tionary and four of them are used to cover full lane-width. The service it provides is primarily for roughness and rutting survey. At this time, there is no independent evidence that laser based technologies are able to provide usable data for pavement crack- ing survey and other condition survey. The primary reason is they do not provide high enough resolution. PAVEMENT ROUGHNESS Engineers have long considered ride quality, sometimes referred to as roughness and sometimes as smoothness, as a favorite indication of the functional capability of a pavement surface. One primary reason is that numerous studies have shown that ride quality is the pavement feature that will most often trigger public response and is a factor in deter- mining vehicle operating costs. A 1995 National Quality Initiative (NQI) survey of highway users found that less than one-half were satisfied with pavement condition, especially smoothness (27). The first 50 years or so of roughness measurement in the United States was summarized in the Background section of this report. It was seen that the technologies applied have evolved from subjective seat-of-the pants methods through response-type road meters to the sophisticated noncontact sensor-based methods in use today. Procedures in Use In the United States, a strong impetus for the collection of roughness data is provided by the FHWA HPMS (5). Although many agencies measured roughness on at least major highways long before the HPMS program was instituted, HPMS re- quired more consistent and uniform measurements. As a result, essentially all highway agencies now measure ride quality of pavements. The HPMS program requires the reporting of IRI for all National Highway System (NHS) roads on a biennial basis. The information from this program is integral to the alloca- tion of federal funds to the states. The HPMS manual cites the following advantages of using the IRI as the roughness measure. It is FIGURE 6 Linescan pavement image with vehicle-cast shadow in left wheel track (Courtesy: Virginia DOT). 4,096 pixels (transversely), 4 meters wide FIGURE 5 A 4,096-pixel resolution image in JPEG format (line scan digital camera) (20).

16 • A time-stable, reproducible mathematical processing of the known profile; • Broadly representative of the effects of roughness on vehicle response and users’ perception over the range of wavelengths of interest, and is thus relevant to the definition of roughness; • A zero-origin scale consistent with the roughness defi- nition; • Compatible with profile measuring equipment available in the U.S. market; • Independent of section length and amenable to simple averaging; and • Consistent with established international standards and able to be related to other roughness measures. The standard accepted by most agencies for determination of the IRI is AASHTO Provisional Standard No. PP37-00 (28). This standard provides for the use of a longitudinal pro- file determined in accordance with ASTM Standard E950, Standard Test Method for Measuring the Longitudinal Profile of Traveled Surfaces with an Accelerometer Established Iner- tial Profiling Reference (29). More than 80% of the agencies responding to the questionnaire reported using some variation of AASHTO, ASTM, HPMS, or World Bank roughness mea- surement protocol. Follow-up phone calls or e-mails to some agencies revealed that in many cases there are different names used for the same AASHTO standard. Other respondents used agency-specific protocols, but provided few details. Sensor Technologies Now, virtually all network-level roughness monitoring re- ported in the United States and Canada is conducted with instrumented vehicles using accelerometers and at least one of three types of sensors: lasers, acoustic, or infrared. The accelerometers provide a horizontal plane of reference, where- as the sensors measure pavement deviations from a horizontal plane. Most sensors work on the basis of a simple concept that the distance from the reference plane to the road surface is directly related to the time it takes for the signal to travel from a transducer to the road and back. Lasers, however, work on the basis of a phase shift of the refracted laser beam in a process beyond the scope of this synthesis (30,31). The faster the signal, the more frequent sampling can be done at a given vehicle speed. Although fairly simple in concept, the measur- ing process is not so simple in application, for very high-speed and high-capacity electronic components are required to cap- ture the large volumes of data generated. The vehicles, known generically as profilers, produce in one pass a “continuous signal or trace related to the true pro- file of the pavement surface” (6). This longitudinal profile is the basic measure of the pavement surface from a ride quality point of view. However, converting profile features into a use- ful index of ride quality was the subject of extensive research that culminated in the almost universally used IRI (7,8). The automated roughness-monitoring technologies in use by the various agencies are summarized in Table 6 by the number of agencies using each type of technology. Lasers are the most popular by a wide margin, and in one case an agency using acoustics reported plans to move to a laser in a replace- ment vehicle. At present, 38 states, 4 provinces, and 2 FHWA offices use lasers. Each of the automated roughness sensing technologies is discussed briefly in the following paragraphs. Laser Sensors Several vendors use or sell (sometimes both) pavement condi- tion survey equipment where lasers are the principal means of profile measurement (22,25,32). The laser technology has evolved rapidly over about the last decade so that they now operate at very high speeds. Such high speeds permit the col- lection of profile data at intervals of 25 mm (1 in.) or less at vehicle speeds of up to 96 km/h (60 mph) (22). Depending on user needs, vehicles may be configured with from 1 up to numerous (often 37) lasers. One laser can be used to measure the longitudinal pavement profile in one location, whereas any number may be used if there are attempts to capture several longitudinal locations and the transverse profile as well. Acoustic Sensors Acoustic sensors were the basis for some of the earlier non- contact profilers developed. One of the best known of these was developed by the South Dakota DOT and is still referred to as the South Dakota profiler. A typical configuration was of three sensors and two accelerometers on the front bumper. In 1986, only South Dakota was using the device, whereas in 1991 the number had increased to 25 agencies (6). Now, as Table 6 shows, only three agencies reported using the acoustic technology for roughness monitoring (more use it for rut- depth measurements). The change appears to be related to the availability of the high-speed lasers that are not subject to some acoustic sensor problems with obtaining reliable mea- surements on coarse textured pavements. Infrared Sensors Infrared sensors are used by a few agencies, principally those using a newer version of the K.J. Law, Inc., Profilometer. That company has recently been sold and the buyer (Dynatest Con- sulting, Inc.) has moved to a laser sensor technology (33). Agency Type Laser Acoustic Infrared State Province FHWA Total 38 4 2 44 2 1 — 3 2 2 — 4 TABLE 6 TECHNOLOGIES USED IN ROUGHNESS MEASUREMENT (Number of Agencies)

17 Few specifications were readily available on infrared sensors at the time this synthesis was developed. Profile and IRI The output from pavement profilers is a standardized process (29) as is the determination of IRI from that profile (34). The first of these, Standard Test Method for Measuring the Lon- gitudinal Profile of Traveled Surfaces with an Accelerome- ter Established Inertial Profiling Reference (29), covers the measurement and recording of profile data. The standard also addresses requirements of profile measurement equipment, the recoding of profile data, calibration requirements, deal- ing with faulty tests, determination of precision and bias, and reporting of the data. The IRI standard, Standard Practice for Computing Inter- national Roughness Index of Roads from Longitudinal Pro- file Measurements (34), covers the mathematical processing of the road profile data to produce the IRI statistic. In addi- tion, the standard addresses the determination of precision and bias as well as reporting of the data. The ASTM profiling standard is also an integral part of the AASHTO provisional standard on IRI. AASHTO, how- ever, refers back to the basic research (8) rather than to ASTM for the actual IRI calculation. RUT-DEPTH MEASUREMENTS Forty-six of the reporting agencies collect automated rut- depth measurements, nearly always concurrently with rough- ness monitoring, because generally laser or acoustic tech- nologies mounted on the same vehicle are applied. A few states reported that their rut measurements are received as a part of the profile measuring program, but that no use is made of the data. Procedures in Use Currently, there is no consensus concerning the number of sen- sors devoted to rut-depth measurement. As can be seen in Table 7, the states are equally divided over the 3- and 5-sensor methods at 16 each, whereas 11 have adopted the “rut bar” with from 7 to 37 sensors (five states use a 37-sensor rut bar). Two Canadian provinces use 37-sensor rut bars and three use 4 to 5 sensors. Eastern federal lands use more than five sen- sors, whereas the LTPP program uses a unique shadow line projection method supplemented by laser profiling. The LTPP method, where 35-mm films are used to capture pavement images, called the PASCO RoadRecon, is described by LTPP (35): The PASCO RoadRecon system incorporates a van driven across the test section at night. A boom, on which a 35-mm camera has been mounted, extends from the rear of the van at the top of the unit. A strobe projector, mounted on the bumper, contains a glass plate that has a hairline etched onto it. The strobe and the camera are synchronized so that when the cam- era is triggered to take a picture, the strobe projects a shadow of the hairline onto the pavement surface at a specific angle in relationship to the van (and thus at an approximate angle to the pavement surface). The coordinates along the hairline image for each picture are later digitized and stored on a computer. Photographs are taken approximately every 15.2 m (50 ft). In practice, few agencies actually use all of the sensors theoretically available on the 37-sensor rut bar. That bar pro- vides for a sensor each 100 mm (4 in.) on a 3.6-m (12-ft) wide lane. However, because of safety concerns, agencies usually limit the bar length to approximately 3 m (10 ft) with 31 sensors. The AASHTO provisional standard on rut-depth mea- surements is given in Figure 7 (36). The equations for calcu- lation follow the figure. This standard is intended for use with a vehicle traveling over the pavement at highway speeds, although it is adaptable to a manual measurement as well. Most of the 15 agencies reporting a five-sensor measurement have adopted some form of this protocol. From the standard comes the following explanation: The transverse profile is determined on the basis of the vertical distance between an imaginary string line run across the traffic lane from the shoulder to the lane line. The string line may bend at the hump between the wheel paths where the hump is higher than the outside and centerline of the lane. For manual mea- surements, the use of a string line will require D1, D3, and D5 to be zero (36 ). Five-Point Rut-Depth Calculation where Ro = rut depth outside wheel path estimate (mm); Ri = rut depth inside wheel path estimate (mm); R D D D R D D D o i = − + = − + 2 1 3 4 3 5 2 2 1( ) Agency Type 3 Sensors 5 Sensors >5 Sensors (Usually 31) State Province FHWA Total 16 0 0 16 13 3* 0 16 11 2 1 14 *Two use five sensors, one uses four. TABLE 7 METHODS OF RUT-DEPTH MEASUREMENT (Number of Agencies)

18 Ro = is not less than 0; Ri = is not less than 0; and D1, D2 . . . D5 = height measured as shown in Fig- ure 7 (mm). Recent Developments Rut-depth measurement has been the subject of much discus- sion in recent literature. The three-sensor method has come under a great deal of criticism, with a number of researchers and others taking the position that it does not measure rutting with sufficient accuracy to be useful (37). The AASHTO pro- visional standard, as described, requires a minimum of five sen- sors. Still others are not satisfied with the five-sensor approach. Simpson (38) has reported that a minimum of nine sensors is required to achieve rut measurements with sufficient accuracy for pavement management. That work was based on compar- ison of a five-sensor rut bar to rod and level elevations and by systematically reducing the number of data points (sensors) that would be required to achieve acceptable correlation co- efficients with “true” profiles. Questionnaire results show that the Vermont Agency of Transportation (VAOT) has adopted a nine-sensor rut bar on a state-owned van, although its con- tractor (Roadware) uses 30 sensors. Then, as given in Table 7, some agencies have gone to 30-sensor rut bars. This approach was introduced in the configuration of the Automatic Road Analyzer (ARAN) SmartBar (39). The sensors are placed 100 mm (4 in.) apart and theoretically provide full lane width coverage. The numerous data points offered by this technol- ogy suggest a much more accurate profile, but no independent validation has been discovered to date. Faced with the fact that the 37-sensor configuration results in a dangerously wide 3.7 m (12 ft) rut bar on the survey vehicle, vendors have introduced still another tech- nology of rut-depth measurement referred to as the scan- ning laser (39–41). An example is depicted in Figure 8. The approach provides for high-powered pulse lasers mounted on the back of the van to project a line across the pavement. One vendor using two lasers claims “lateral resolution is 1280 points across the width of the pavement (4 m). Depth accuracy is 0.5 mm or approximately 1⁄32 in.” (39). The same vendor continues, “Proprietary software will enable the vendor to provide the same 37 points of data to permit the use of existing well-proven algorithms and to provide com- patibility with existing customer databases” (39). This tech- nology was developed by a Canadian optical technology firm (40) and no independent validation has been pub- lished. A different company (41) developed the one laser system depicted in Figure 8 and, again, no independent val- idation is available. JOINT-FAULTING MEASUREMENTS Rigid pavement joint faulting is a distress feature that many agencies do not collect, in part because they do not have FIGURE 8 Scanning laser [Courtesy: Mandli Communications (41)]. 2300 mm (min.) Length set by user Centerline Inside Wheelpath Outside Wheelpath Reference Plans D5 D4 D3 D2 D1 FIGURE 7 Rut-depth measurements [after AASHTO (36)].

19 INTEGRATED SYSTEMS The emphasis on the collection of massive amounts of pave- ment condition and other roadway data over the past few years has led to a proliferation in the development of what can be described as integrated systems. In a single pass of a data collection vehicle configured as an integrated system, a variety of data will be collected, depending on specific ven- dor and user requirements. Some of the data elements col- lected are pavement, right-of-way, and other images; longi- tudinal and transverse profile measurements; and in some cases, a texture-measuring system. All incorporate a distance measuring instrument (DMI), while many now include a satellite-driven GPS. Although many systems are vendor owned and operated, others are sold directly to user agencies. An example of such a multipurpose vehicle is given in Figure 10. This van features a digital line scan camera mounted above a very intense light- ing system. In questionnaire responses, several agencies indicated that they were in the process of buying or upgrading systems. In many cases, one emphasis will be on achieving real-time cracking, smoothness (IRI), rutting, faulting, and other data in P1 and P2 Points to Measure Relative Elevation 300** **75 to 225 **75 to 225 P1 P2 Approach Slab Subbase or Subgrade Transverse Joint Departure Slab * Variable ** Constant FIGURE 9 Points to measure for faulting by automated measurements (all measurements in millimeters) [after AASHTO (43)]. jointed concrete pavements. Of those that do collect the data, seven use manual or visual methods. Several agencies do a visual assessment as part of or in addition to overall visual condition surveys. Where joint faulting is measured manu- ally, most agencies use specially designed faultmeters, the most popular of which is the Georgia Faultmeter developed by the Georgia DOT (42). A few agencies do some manual measurements by agency personnel, while vendor data are collected through automated means. Procedures in Use Twenty-three of the reporting agencies apply automated means to collect joint-faulting data. Those agencies over- whelmingly combine faulting measurements with roughness monitoring, because IRI is also affected by joint faulting. Most use two sensors (usually to measure longitudinal pro- file as well) and apply protocols developed by vendors or equipment manufacturers. Table 8 summarizes the methods reported in the survey responses. The AASHTO provisional standard for joint-faulting mea- surements provides for the methodology given in Figure 9 (43). It is applicable to both manual and automated methods. Here, the user may vary the distance from the joint to the points of measurement, but must keep those points 300 mm (12 in.) apart. Faulting is defined as simply the elevation difference between the two points of measurement (P1 and P2) to the nearest 1 mm (0.04 in.), with a difference of 5 mm (0.2 in.) defined as the threshold of faulting. In questionnaire responses, only four agencies cited the AASHTO or a modified AASHTO standard. Most listed an agency or vendor protocol and provided few or no details. Agency Type Manual Sensor State Province FHWA Total 9 1 1 11 22 1 — 23 TABLE 8 METHODS OF JOINT-FAULTING MEASUREMENT (Number of Agencies)

20 ber of pieces of equipment in use, but rather the number of agencies to which a given supplier has either sold equipment or has provided recent hired services. Finally, Texas manu- factures the equipment it uses. Retired equipment or older hired services are not reflected. Each of the commercial equipment suppliers listed in Table 9 has a corresponding World Wide Web address listed in the website directory in this synthesis. In most cases, detailed descriptions of the equipment furnished are provided in the various websites. SUMMARY It is clear from the responses to the survey questionnaire and discussion in this chapter that the automated collection of pavement distress data is in a state of rapid evolution. This finding probably is most true in the area of cracking or surface distress collection, where the imaging technology may be changing faster than the users can adjust to those changes. This situation is highlighted by the change in just one decade from the use of 35-mm film to capture most images through the use of analog video technologies to the use now of high-speed, high-resolution digital cameras and line scanning instrumen- tation. Still, some agencies reported that they are struggling with achieving pavement images at highway speeds and with acceptable quality. Roughness monitoring is a more mature science wherein the major changes over the past few years have been in the pro- tocols applied and in the new technologies brought into use, primarily in the area of the sensors used. Currently, high-speed laser sensors have replaced practically all of the acoustic and infrared types for longitudinal profile measurement and the resulting IRI statistic. Automated rut-depth measurement constitutes a science that is in an unsettled state, because there is so little consensus on the number of points to measure (how many sensors to use). Although an AASHTO provisional standard suggests a mini- mum of five sensors, many agencies still use three. However, one researcher is promoting at least 9, whereas some of the vendors have moved to 30 or more, and some are promot- a single pass of a data collection vehicle. A typical new inte- grated system will incorporate one or more versions of imag- ing equipment, an ASTM Class II profiling system, and a means of positive location reference. Many systems will have the ability to provide roadway lighting synchronized with imaging equipment to alleviate the problem of shadows from either the equipment itself or from features surrounding the roadway. Added to some systems are right-of-way images and methodologies to provide such information as guardrail and signage inventories. Gunaratne et al. (44), at the University of South Florida, have reported on a comprehensive evaluation of a vehicle proposed for pavement surveys by the Florida DOT (FDOT). This ICC vehicle contains digital imaging, laser sensing, and location-reference technologies and was shown by the researchers to be capable of fully meeting FDOT’s pavement data collection needs. However, it was noted that visual sur- veys would still be needed “until FDOT acquires reliable soft- ware for automatic analysis of pavement distress videologs.” A paper offered for presentation at the 2004 TRB annual meet- ing continued the evaluation, focused on digital images from the Florida vehicle, and identified several limitations of the equipment (45). Among these was a finding that light “noise” is very difficult to cope with in daylight imaging therefore night operations were recommended. Table 9 lists the systems in use, by manufacturer, as given by the responses from the various agencies. In that table, there is no differentiation of agency-owned and hired equipment. There is also no claim that the listing is complete, because some agencies were hesitant to mention equipment manufac- turers’ names for various reasons. Furthermore, the table does not reflect that many agencies where the equipment is owned will own a number of the same devices. One agency reported owning eight profilers by one supplier. In addition, several agencies own equipment furnished by more than one of the suppliers listed. Therefore, Table 9 does not reflect the num- Supplier Agencies Using Dynatest and Law GIE Technologies International Cybernetics Corporation (ICC) INO Pasco/CGH/ERES Pathway Services Roadware Group, Inc. Agency Manufactured 5 2 9 2 1 9 19 1 TABLE 9 EQUIPMENT IN USE FIGURE 10 Typical multipurpose van [Courtesy: International Cybernetics (25)].

21 ing thousands of data points collected by a scanning laser. Rut-depth measurement clearly is an area needing additional research to arrive at an optimum testing scheme and protocols. Automated joint-faulting measurement is an area that has not received great emphasis by many agencies. This finding is demonstrated by the relatively few agencies actually collect- ing the data through automated means. Some of those that do use automated means professed to have little confidence in the data collected. If others collect the data at all, most are using manual methods such as simple straight edges or the Georgia faultmeter. Although the latter seems to have a good reputa- tion among users, its greatest disadvantage is that it is still manual, extremely time consuming, and limited to project- level work. Again, there is a need for further investigation of automated methods of joint-faulting measurements. Equipment manufacturers and vendors are continually up- grading pavement condition data collection and processing equipment to incorporate the latest technologies. Much of this effort is inspired by a desire for more real-time data analysis as more agencies collect more data. Almost every vendor maintains a website with descriptions of the equipment avail- ability and features.