| Along with the enhancement of China’s comprehensive national strength, the rapid development of the highway has become a necessity, the pavement maintenance is getting more and more attention. However, the manual work efficiency is low, the accuracy is also difficult to be guaranteed, the vehicle type road surface defect detection system arises at the historic moment.In the actual work, the pavement defect detection will encounter all kinds of complex scene, such as: pit, oil, shadow, road signs and other factors, to the detection of a great difficulty, the existing algorithm is difficult to ensure the efficiency and accuracy, often lead to the need for a lot of repair is ignored, the road defect detection algorithm has become a technical bottleneck. In view of the above problems, this paper makes a deep research on the pretreatment of the pavement image, the information extraction of the surface defect, the intelligent repair algorithm and the design of the system. This paper mainly completes the following work:1, the collection of road surface defect image preprocessing. In view of the uneven brightness of the road surface defect images, the gray level correction is carried out to improve the image brightness and contrast of the glare caused by the glare. After gray scale correction, this paper adopts an improved adaptive filtering algorithm, which can effectively suppress the noise of the pavement crack image, and preserve the details of pavement cracks.2, pavement crack information extraction. After preprocessing, image segmentation is carried out, and a method based on non negative feature extraction is proposed, which is based on the image feature of the background, and uses the method of connected domain based on background label to quickly traverse the crack image and reduce the time of processing of the image processing. Then the crack skeleton is extracted by morphological transformation, which makes the image more simple and intuitive. Then, the chain code tracking and extraction technique is used to calculate the parameters of the crack in order to describe the shape, then use the area and the length of the threshold method to remove the short chain and better express the crack information.Experimental data verify the effectiveness and generality of the proposed algorithm. In the experiment, to ensure the correctness of the algorithm, each step is verified, and then the current crack extraction algorithm to do a comparison. The experimental results show that the crack detection and recognition algorithm breaks through the bottleneck of the existing extraction algorithm. The algorithm has high sensitivity and strong anti noise edge characteristics. It can extract the target information in various complex pavement images accurately. |
