Research On Pavement Distress Detection And Recognition Based On Computer Vision

With the rapid development of highway construction in our country, the detection and maintenance of highway has been paid more and more attention in the construction of national economy and people’s livelihood. The traditional manual testing can not meet the requirements of the rapid development of highway.Therefore, the research on automatic detection technology of pavement distress becomes more and more important. In recent years, the pavement distress detection system based on computer vision has been widely studied in the field of highway maintenance.This paper focuses on the hot spots and difficulties in the pavement damage detection algorithm.Firstly, there are some problems such as uneven illumination, serious noise interference and so on. In this paper, the image pretreatment is proposed, median filter is used in the process of image removing noise, the results are compared with mean filter and Gauss filter. The experimental results show that median filter can get better results in removing noise points.Secondly, edge detection and region segmentation. A pavement crack detection algorithm based on dual-tree complex wavelet transform(DTCWT) was proposed in this paper, combining direction gradient histogram with dual-tree complex wavelet transform. First, this algorithm decomposed pavement crack image into six sub-band images with dual-tree complex wavelet transform. Then, calculated the histogram direction gradient matrix of each sub-band image. Threshold processing could be used to determine the edge of the crack. Proved by experiments, compared with the traditional edge detection algorithms, the algorithm proposed in this paper had high target recognition degree, strong anti-interference ability, and high accuracy rate.After the crack edge is determined, the area is filled, and the crack area is separated completely.Finally, the feature extraction and classification are studied. In this paper, the feature extraction of pavement crack image after threshold, the support vector machine classifier is designed according to the extracted features. Recognition and classification of pavement images. Through 90 pavement crack image test, the classifier can be used to classify pavement cracks effectively.