Road marking assessment using digital image analysis

Abstract

This paper presents a systematic approach for the automated measurement of road marking wear (erosion) using digital video image analysis techniques. These techniques are used to digitize an image of the road surface and thereafter manipulate the image to obtain the extent of any road marking within the image. Simply, after digitization, a process known as image segmentation was used to analyse the image to ascertain the picture elements (pixels) within the image that were clearly road markings and those which were not. Subsequently, a further process, called image enhancement, was developed to remove misclassified pixels. Finally, a novel process (feature description) was used to compare the road marking with an ‘ideal’ to determine any erosion of the original road marking. The above system was tested on a set of images of road markings. The accuracy of this system in comparison to a manual method of assessing road marking erosion was investigated using statistical techniques. The results show clearly that the automated methodology is accurate and reliable.