Crack Detection on Asphalt Surface Image Using Enhanced Grid Cell Analysis

This paper presents image processing techniques based on grid cell analysis for crack detection on non-uniform illumination and strong texture images. The techniques are grid cell analysis chain and cracked cell verification. Grid cell analysis chain helps eliminate false detection of shadow border on an image as being a cracked line. Cracked cell verification helps eliminate false detection when a grid cell has some image noise or strong texture but does not really contain part of a cracked line. Good accuracy in finding cracked lines on a pavement image with non-uniform illumination and strong texture can be achieved with 13% and 21% of false positive and false negative respectively.