Research On Asphalt Pavement Crack Detection Algorithm

The highly effective and reliable road crack automatic detection and recognition system has a significant effect on traffic safety. A high level of road crack detection and recognition system become one of the frontier research topics of transportation field, for the traditional road detection technology can not meet the requirement of real-time and efficient.Besides analyzing the detection of the pavement from home and abroad, the key technologies about automatic detection and identification of pavement crack was studied in the paper. Image preprocessing, image binarization and morphological processing were used to deal with the asphalt surface image in order to judge whether the cracks existed in the process of the detection and identification. If the cracks existed, we extract the crack shape and calculate the crack length, width and other geometrical features. The main tasks are as follows:(1) According to experiment and analyzing of several common filtering methods, median filter method and Gaussian filter method were improved in this paper. Experimental results show that the Gaussian function filtering method has a better effect on filtering of asphalt pavement image than other method.(2) Binarization method was used to deal with the image of the asphalt pavement after filtering, and then the results were analyzed. Two block-based binarization algorithm were designed which were based on the common binarization method's shortcomings. These two algorithms were subject to local minimum gray value of crack in asphalt pavement. Weighted difference and differential phase shift method were used to separate the background and crack pavement.(3) Pseudo-target was eliminated by closing operation method and opening operation method from image after binarization. Then, using the method of connected component labeling, skeletonization, calculation length and width of crack, logic calculation to deal with the image and a better shape of crack was extracted.The actual asphalt pavement image processing results showed that pavement crack detection and identification was completed by the good result in this paper. Comparing this result to other similar systems, this paper's results not only appear crack loss, but also have a smaller error with real facts.