As part of the development and implementation programme to support the introduction of SCANNER surveys on all local roads in England, this project investigated the capabilities of crack detection systems on all types and classes of local road, from principal roads to minor unclassified roads.
This research was carried out by TRL Ltd in two phases. Phase 1 of their work is complete and the interim report is available here: Crack detection on local roads - phase 1
Measurement of the extent and severity of cracking is a cornerstone of visual inspection surveys and is often the trigger for maintenance. However the automatic detection of cracking from images is fraught with difficulty and there was a need to improve the performance of the systems currently available in the UK. In the first phase TRL reviewed the requirements of local authorities for crack detection and measurement and investigated the performance of currently available systems that had the potential for further improvement or enhancement to meet engineer's requirements.
Following a review of all the different methods of crack detection, and the systems available worldwide, a comparative trial was carried out on 5 different systems from Canada, England, Sweden and the USA. None was found to be comprehensively superior to the others, and all had similar difficulties excluding a range of surface features commonly found on British local roads from the crack detection and analysis.
Consequently TRL recommended that the second phase should concentrate on three areas. Investigate generic methods of improving the detection and elimination of distracting features such as joints, the edge of patches, fretting, ironwork, high friction surfaces, road markings and the edge of carriageway, improve the methods of using crack data in the assessment of roads at the network and local levels, and modify the acceptance test procedures.
The second phase of the research has been completed and the final report is available here: Crack detection on local roads - phase 2
The second phase concluded that control of image quality is essential if subsequent image processing algorithms are to be effective, and this can be improved both through hardware developments (control of illumination and camera quality and positioning) and software (processing the images to achieve consistency).
A number of different techniques were developed and demonstrated to identify and filter out non-crack features that give rise to "false positive" reports of cracking, and to classify different types of crack. The final report also recommends improvements in the specification for crack measurement and the acceptance test procedures.
These results have been implemented in the SCANNER User Guide and Specification for 2007.