Research on Road Crack Detection based on Machine Vision

Road cracks will damage the pavement structure, seriously affect traffic safety and driving comfort. Aiming at the deficiency of manual detection, in order to improve the efficiency of crack detection, machine vision technology and image processing method are used to detect road cracks and extract crack parameter information. Firstly, the collected crack image is processed by graying, smoothing and grayscale stretching. Then threshold segmentation is carried out, and the background noise is removed to extract the crack target. Finally, according to the statistical curve change trend of the number of crack target pixels from different directions for different types of cracks, the type of cracks is judged; The length of the crack is calculated by extracting and refining the skeleton of the crack, and the average width, maximum and minimum width of the crack are calculated. The area is calculated by the minimum circumscribed rectangle. The test results of the detection system show that the accuracy of crack detection is basically more than 90%, the accuracy of crack area, length and average width are more than 95%, and the accuracy of crack classification is 100%.