Research On Pavement Crack Detection System Based On Image Processing

Looking at the growing scale of highway construction in our country,the demand for the use of pavement is increasing,which causes the serious damage to the pavement.Therefore,how to manage the highway scientifically and maintain its normal use status is the primary challenge in the field of highway construction.The damage of roads should be discovered timely,only in this way we can improve its service life.With the continuous development of computer vision technology,image processing has become a hot research topic nowadays.Image detection and processing technology are introduced into pavement crack detection system.There are some influencing factors such as the irregular texture of road image,the weak signal strength of detection system,which make the recognition of the roads difficult.Based on the above problems,this paper studies the road pavement crack detection system,and applies image enhancement processing,image edge detection,edge connection and image feature recognition technology to pavement crack image detection.First of all,in the process of image enhancement,an improved algorithm based on fuzzy theory is proposed to enhance the crack image and it is proved that this algorithm enhances the contrast between the background information and the crack target through the simulation experiment.Secondly,an improved algorithm based on wavelet transform is proposed in the process of image edge detection,compared with the 8-direction sobel algorithm and the improved Laplacian algorithm,experiments show that outperforms of the 8-direction sobel algorithm is better than the improved Laplacian algorithm.Compared the improved wavelet transform algorithm with the 8-direction sobel algorithm,this experiment shows that the peak signal-noise ratio is improved by 6dB.The structural similarity is increased by 50 percent,and the error rate of image detection is decreased by about 9 percent.And we also present a new canny-based algorithm for the connection of discontinuous edge points in the image after edge detection.By combining with the adaptive double-threshold algorithm,the discontinuous edge points are successfully connected and the correct connection rate of crack is increased to about 90 percent.Finally,the fracture features are extracted based on the projection and damage factor algorithm respectively,and the cracks are identified according to the differences of the eigenvalue.