ABSTRACT
Pavement evaluation is a critical practice for the maintenance of roadway transport networks. Nevertheless, the current methods for pavement defects detection and assessment are subjective, costly and time-consuming; creating the need for automation of the underlying processes and for the use of low-cost technologies. In this paper, an automated system is presented for the identification and quantification of pavement patches; a crucial part of pavement surface assessment and rating, which depend on the proportion of road segments covered by patches. The proposed vision-based algorithm uses road surface frames collected by a camera, located on a vehicle moving in a real-life urban network. A Support Vector Machine is trained and tested by feature vectors, generated from the histogram and two texture descriptors of non-overlapped square blocks, which are located in “patch” and “no-patch” areas of the collected images. The outcome is composed of block-based categorization, image-based classification, and measurement of the patch area.
