Image processing technique in automated pavement evaluation system

Date of Completion

January 1998


Engineering, Civil|Artificial Intelligence




Highway systems continuously deteriorate due to exposure to traffic loading and environmental factors. In order to keep highways in good condition, maintenance and repair strategies must be based on an informed knowledge of current pavement conditions. Visual observation of pavement distress is the most common method for monitoring and evaluating pavement surface conditions. This has been traditionally performed by trained engineers who walk or drive along the road and count the distresses. However, this visual survey method takes too much time and effort, is too costly, and is often dangerous due to exposure to traffic.^ Advances in video technology made it possible to capture and record pavement images on high quality media to create a permanent visual record of the actual pavement condition. As a result, this new video technology eliminated thousands of field trips by engineers so that the time and cost needed to evaluate pavement condition has been significantly reduced. However, even with this new technology, a great deal of repetitive and tedious effort was still required since engineers must view all of the recorded pavement images in order to identify and count distress conditions. Therefore, there was a need to develop an image processing method to automatically analyze the recorded images. The goal of this work was the development of an automated and integrated system for the evaluation of pavement distress condition. A new image segmentation algorithm was developed to isolate distress features from images (such as the ConnDOT images) that had excessive noise and significant variation in contrast level. Logistic regression screening was also used to minimize problems with images that contained abrupt and extreme changes in contrast level.^ The results from the system showed a 89.7 percent of accuracy in predicting the absence or presence of distresses. The automated pavement evaluation system developed from this research reduces the overall cost of conducting pavement distress evaluation. The increase in efficiency from converting to an automated system also allows pavement management personnel to more easily develop accurate and comprehensive databases for pavement management. ^