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.

56 Questionnaire responses in conjunction with the related liter- ature permitted the identification of several problems common to a number of the agencies practicing automated distress data collection. Although not all problems are serious, all have caused at least some loss of productivity in data capture or pro- cessing, or some compromise in the quality of data collected or in the final results obtained from collection and processing combined. The whole process of automated distress data reduction from images is evolving and is extremely complex, with sig- nificant technical demands, from the points of view of both equipment and personnel. Major issues have been the inabil- ity to reliably identify certain distresses, the resolution of cracking (the size of crack identifiable), and the rate of pro- duction of high-quality information. As well, real-time, on- board, distress data processing often cannot be carried out at anywhere near the speeds of prevailing traffic, although the collection process can take place at those traffic speeds. Shadows on images are a significant problem for both manual and automated processing of surface distresses from images. Shadows make it very difficult to discern distresses in the shadowed area, and if such areas are large and frequent, they can significantly affect the results of both extent and severity determinations for a given roadway. Some agencies and vendors use lighted cameras to overcome the shadowing problem, especially on rural roads with narrow rights-of-way and where surrounding foliage and other factors are a prob- lem. Furthermore, the survey vehicle and camera layout must be configured properly to minimize shadows from the vehicle itself. Even under the best of other conditions, the position of the sun may make satisfactory imaging without strobe lights almost impossible. There is very little consensus on the methodology of auto- mated rut-depth measurement except that the more sensors used the better the results. The AASHTO provisional standard provides for a minimum of five sensors, although a number of agencies still use three. On the other hand, one researcher is promoting at least nine. Some of the vendors have moved to 37 sensors, whereas some are promoting more than 1,000 data points collected by a scanning laser. Automated rut-depth measurement 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 been addressed by many agencies. That is demonstrated by the relatively few agencies actually collecting the data through automated means. Some of those that do use automated means profess to have little confidence in the data collected. If others collect the data at all, most are using manual methods, such as simple straightedges or the Georgia Faultmeter. Although the faultmeter appears to have a good reputation among users, its greatest disadvantages are that it is still manual, extremely time consuming, and primarily limited to project-level work. Again, there is need for a serious look at how to proceed with automated joint-faulting measurements. Closely related to all of the perceived problems are the AASHTO Provisional Standards on Pavement Data Collec- tion. There appears to be sufficient adoption of the roughness measurement standard and considerable interest in the one on rut-depth measurement. In the latter case, the community needs to decide what is needed before a realistic standard can be devised. Both the cracking and joint-faulting standards are similar to that for rut depth, for there is not much agreement on what is needed. However, the cracking standard has an additional problem in that agencies have existing databases built on agency protocols and comprising pavement perfor- mance histories. Because the agencies often do not want to abolish the old databases they are reluctant to adopt a new protocol. A few are incorporating the AASHTO standards into agency standards. Several areas in the field of automated distress data col- lection could be the subject of further research. First, with the exception of the International Roughness Index, there is clearly a need for further work with data collec- tion protocols. The current generation of cracking, rut-depth measurement, and joint-faulting standards was developed pri- marily for a manual testing environment and later revised to apply to automated data collection. User agencies are not read- ily adopting those standards. A study should examine AASHTO, ASTM, and other stan- dards to determine their applicability to the rapidly evolving field of automated collection. Such a study could further exam- ine the reasons that agencies are hesitant to adopt existing stan- dards and identify what changes would promote adoption. The impact on legacy data would be expected to be a major issue in any changes. The expected results would include recom- mendations of necessary revisions, including the means of CHAPTER TEN CONCLUSIONS

57 consolidating various standards. Another element of data col- lection standards needing some attention is in the area of port- land cement concrete pavements. There appear to be few stan- dards in use for those pavements, although they constitute a significant portion of the highway inventory. Several methods of reducing pavement surface distress (principally cracking and patching) from analog or digital images are currently in use. These have been developed and are being implemented by industry or academia. A research study could examine the relevant technologies, identify the features needing standardization, and provide recommenda- tions on the development of appropriate standards. Processing standards themselves would be developed under a separate project as needed. In addition, survey results show that there is an urgent need for the development of quality management programs for automated pavement data collection and processing. Some agencies, especially those in Canada, have made significant progress in developing quality management programs and procedures, whereas others depend totally on vendors to pro- vide the quality of data needed. That level of quality itself is not well defined except in very general terms, such as requir- ing that the data to be of sufficient quality to feed pavement management program algorithms. TRB Committee A2B06, Pavement Monitoring, Evalua- tion and Data Storage, has recognized the need for a research project to address data quality management. The committee developed a research needs statement with the objective of establishing guidelines that highway agencies can use to develop or improve their quality management practices for contract pavement distress data collection. The guidelines should include the following: • Recommendations on appropriate levels of accuracy, precision, and resolution, as well as how an agency can determine these values for its local pavement manage- ment system decision process; • Required elements of a contractor’s quality control plan; • How to structure the highway agency’s quality assur- ance program; • Ways to evaluate and select contractors; • Appropriate uses of qualification test sections for con- tractor selection, certification, and quality assurance; • How to compare and evaluate vendor test measurements against reference values; • How to measure contractor performance; • Appropriate award and penalty structures for contractor performance; • Progressive data delivery schedules to permit ongoing assessment during the contract period so that corrections can be made; • When to terminate a contract for poor performance; • The magnitude of agency resources required for alter- nate quality management approaches; and • Pitfalls to avoid in development of requests for con- tractor services. Additional study could address the inherent variabilities of the processes, including the sources of those variabilities. There are several critical elements that could be captured and quantified to facilitate the development of better standards and contract provisions including: • Surface distress measurement variability—How repeat- able are the various manual, semi-automated, and auto- mated means of data reduction? • How repeatable are the sensor measurements of rut depth and joint faulting? • What are the sources of variability of such aspects? For example, with image-collected data, how much of the variability is associated with imaging, how much with manual and automated means of data reduction, and how much with data manipulation and computation? Also, are there other sources of variability? If image- reduced data are to be compared with manually col- lected data, what are the variability properties of those data and how well should the two types of data com- pare? Similar questions could be asked for the sensor- collected data. Agencies wishing to use automated data collection and processing have many choices to review and decisions to make. Development of a “toolbox” addressing those choices is an important research need. Such a toolbox would address the various issues, what factors to consider, and the poten- tial tools to address those issues. It is expected that the tool- box would contain information on hardware, software, and procedures for both the collection and processing of pave- ment condition data, including that collected from images and from sensors. The toolbox would also be expected to contain information on the establishment and implementa- tion of rudimentary data quality management of pavement conditions. In addition, there is an overall need for the development of standards for pavement distress data. Although there are numerous data formats and handling procedures in use in the pavement data collection arena, nothing has been standard- ized. This means that essentially every vendor can use dif- ferent formats and procedures and that customers and others can be at a loss as to how a given data system works. Stan- dards would offer another tool that agencies and others could use in the development of data procurement contracts. It would be much easier to specify that a data management sys- tem meet a certain standard than as is now done with spelling out specific formats, equipment, etc. Another important jus- tification for data standards is to provide for data documen-

tation to ensure against the loss of data utility whenever key personnel are no longer available. Finally, the general term applied to information about data is “metadata,” or data about data. Simply put, metadata are the background information that describe the content, quality, condition, and other appropriate characteristics of 58 the data. It is expected that the standards developed would address at least data format, data storage, data access, data transfer, and data documentation. It is further expected that the standards would address data at both the systemwide and elementary levels. Examples are analysis models for crack detection and classification, and linear referencing of image location, respectively.