Research On Image Detection Algorithm Of Road Surface Damage

With the rapid development of the transportation industry, the road network has been improved, highway maintenance and conservation have been received more and more attention. Traditional manual detection and identification methods cannot meet the requirements of today’s technical requirements because the great demand of the road automatic detection in highway fastpaced development. So based on the pavement image detection and recognition technology become particularly important. This paper focuses on the pavement damage image of crack detection technology.Firstly, this paper analyzes the characteristics of the image, and preprocessing it. The pavement damage image introduced lots of noise inevitably since road conditions, collection devices or other reasons. This article uses the method of Histogram equalization to enhance the image, uses Wiener filter to filter the image, and contrast with the effect of Median filter and Average filter, draw the conclusion that Wiener filtering is better in the pavement damage image of crack detection and has a certain effective in the subsequent processing.Secondly, after preprocessing, we need to study the problem of edge crack detection and segment. This article use Sobel operator to detect the pavement damage image, and get an ideal effect. This paper using Binary morphology algorithm to deal with the image, eliminate the isolated noise point and small area regional noise in the image. Through this, not only can keep the clearness and completeness of the edge goal, but also have a better effect on controlling the noise, and use morphology algorithm to deal the edge crack, which in turn realize the fracture region segmentation.Last, we studied the feature extraction and classification problems. This paper extract three features of pavement crack, they are as follows, cracks image of projection features optimized, statistical the crack area sizes, the crack of distribution density. This paper use the three features then we extracted which the SVM algorithm to classification the pavement damage image. Through the study of the experiment of 176 images, we draw the conclusion that use the three feature which we mentioned can classify the crack image well, the projection characteristics after optimized make more prominent of the crack characteristics, more effective of the resistance to noise, higher identification accuracy, smaller detect error.