Research And Implement Of Vehicle-mounted Automatic Pavement Distress Detection System Based On Line-scan Camera

With the rapid construction of Expressway in our country, the pavement distress detection becomes a very important issue. Currently, semi-automatic detection vehicle is widely used in pavement distress detection. This kind of equipment still needs manual to handling the offline image data and can’t achieve the full automatic detection. Having concerned on the pavement detection requirement and situation, we carried out the research and implement of vehicle-mounted automatic pavement distress detection system based on line-scan camera.In this dissertation, we use two line-scan cameras, image acquisition card, and combined with LED light source, IPC, encoders and other auxiliary devices to achieve the acquisition and storage of the pavement image data. The whole system is integrated and installed in a commercial vehicle. We designed and verified a full automatic image processing algorithm for the pavement distress detection and build a software platform which the detection accuracy can reach more than 90%.The key issues we discussed here are: the adjustment of pavement crack image under non-uniform illumination and the removal of pavement markings, an improved method based on MSR and a combined Hough transform with an improved Sobel operator are proposed; Image filtering and nonlinear gray scale transform taken to highlight the crack information; Then, edge detection, threshold segmentation based on one-dimensional maximum entropy, mathematical morphology method to removal the isolated noise and based on the morphological characteristics to classifiy and recogonition the real cracks in the road. Finally, combined with the machine vision software HALCON and the visual studio development environment developed a set of pavement distress automatic detection system. The experimental results show that the procedure is quickly and accurately, and meets the requirement of design at the beginning of the detection system.