PRIMITIVE-BASED CLASSIFICATION OF PAVEMENT CRACKING IMAGES

Collection and analysis of pavement distress data are receiving attent ion for their potential to improve the quality of information on pavem ent condition. We present an approach for the automated classification of asphalt pavement distresses recorded on video or photographic film . Based on a model that describes the statistical properties of paveme nt images, we develop algorithms for image enhancement, segmentation, and distress classification. Image enhancement is based on subtraction of an ''average'' background: segmentation assigns one of four possib le values to pixels based on their likelihood of belonging to the obje ct. The classification approach proceeds in two steps: in the first st ep, the presence of primitives (building blocks of the various distres ses) is identified, and in the second step, classification of images t o a distress type (using the results from the first step) takes place. The system addresses the following distress types: longitudinal, tran sverse, block, alligator cracking, and plain. Application of the model s to a set of asphalt pavement images gave promising results.