I work for the Florida Dept. of Transportation in the area of Pavement research. We have had the acquisition technology for over 10 years now which is what you were referencing. It involves line scan cameras, IR lighting, GPS, profiling lasers, etc. The problem has always been to process this data automatically. We have seen that the accuracy needed for our purposes always fall short with automated analysis techniques. We have implemented a semi-automated method that allows the user to edit the processed images and correct false positives or add to cracks for areas that were not picked up automatically in the program. Image quality is a big issue too. Our van collects data at 60mph and saves an image that is 3mb in size and 13'X20' of actual roadway. Even with the high resolution cameras false positives are always an issue. We have recently inquired about 3D camera systems that act more as a laser that a camera. They assign a depth value to each pixel on the image, which allows for more accurate data analysis and less false positives. With these systems cracks are being detected even if they are 1-2mm in width. Currently we will use the system we have, and do post processing to achieve the accuracy we need, but in the future we would like to acquire a van with a 3D camera system.
