| In recent years, the highway is developing rapidly, and so the management and maintenance job of the highway also needs to keep up with the development of the construction. At the same time, the quality control system of the highway roadbed and pavement is becoming more complicated, and the request for the measurement is also much stricter. And in our country, the pavement evaluation equipment of the highway management and maintenance department is severely insufficiency, and most of the time, the quality evaluation methods depended on manual method. This method has many disadvantages, such as inefficiency, tiring, slow speed of evaluation and big artificial error etc. And so this situation is highly out of keeping with the development of the highway. Therefore, the researches of the automatic evaluation system of the highway are showing much urgent. It's a complicated system which is involving with the domain of optics, electronics, mechanism and the digital image processing etc. The automatic evaluation of the pavement crack, the orientation of the crack and the system design are three important techniques among this domain. This paper mainly study on the algorithm of the evaluation and recognition of the highway asphaltum pavement crack, and mainly concentrate on the crack image processing and the recognition of the crack type. Therefore, this paper brings forward a system platform of the automatic checking and recognition of the pavement crack image on theoretics. This platform is including with four parts: CMOS digital visions, regulated power supply, lighting system and the compute system. At first, the paper analyzed the features of the pavement crack image before carrying through processing of the crack image. At this rate, we can have an intrinsically understand of the crack image feature, and solve the crack image enhancement and segmentation problem at root. In image enhancement, this paper brings forward an image enhancement method based on fuzzy relation according to the crack image feature after analyzing several image enhancement methods. The principle of image enhancement method based on fuzzy relation is that we firstly transform the image from spatial field into the fuzzy field, and then carry through the image enhancement on fuzzy field; at last we transform the crack image from fuzzy field into spatial field according to the transfer function. After testing this method using actual pavement crack image, it can find that this method has a good enhancement performance which can restrain the noise and eliminate the nonuniformity of background illumination. In image segmentation, this paper adopts a thresholding method based on maximum variance ratio. We can know from the pavement crack image feature that the histogram of pavement crack image takes on "unimodal"feature, the traditional optimum thresholding will be invalidation, for the traditional optimum thresholding is aimed at the "bimodal"histogram image. The thresholding method based on maximum variance ratio has a good performance for "unimodal"histogram image, but it will be invalidation for "bimodal"histogram image. So this paper adopts thresholding method based on maximum variance ratio to segment pavement crack image. Because of the complexity and coverall information of pavement crack image and the difference of the material particle, it make the image segmentation becomes very difficult. Although the thresholding method based on maximum variance ratio has a good performance for pavement crack image segmentation, the binary pavement crack image has many noise after segmentation because for those reasons. And these noises are in the form of "isolated"points or "isolated"blocks. And so these "isolated"points or "isolated"blocks must be eliminated in order to guarantee the accuracy of the checking and recognition of the pavement crack image. This paper brings forward a method based on chain code following to eliminate these "isolated"points or "isolated"blocks. It is shown that this method has a...
|
PDF Full Text Request