Algorithm Research Of Automated Identification Based On Image Engineering For Pavement Distress

Pavement distress causes adverse effects on high-grade highway in many aspects, such as carrying capacity, durability, vehicle speed, mechanical wear, fuel consumption, driving comfortableness, environmental protection, traffic safety, etc. Traditional method based on artificial vision for pavement distress detection has a lot of shortcomings such as time-consuming, dangerous, costly, inefficiency, low accuracy and so on. In recent years, pavement distress detection based on digital image engineering has been developed greatly in highway maintenance field. However the automatic detection algorithm is still not satisfying. Artificial vision method is the main measure for the later image processing. This paper is devoted to research the automatic detection algorithm including pavement surface image enhancement, denoising, image segmentation, crack classification, crack abstract and crack positioning.The main work is described as follows.1. The images which gathered from CCD camera have many defects including non-uniform illumination, severe noise, fuzzy edge of crack and many breakpoint,etc. An improved image denoise algorithm based on contourlet transform is proposed to overcome the drawbacks. By using contourlet transform, the noised image is decomposed into a low frequency subband and a set of multi-sacle and multi-directional high frequency subbands. The high frequency coefficients of the original image are processed by diverse relevance. The noise which has small area is removed, and the small feature which has large or consecutive support area is preserved. The denoise image is gotten by performing the inverse contourlet transform to these estimated coefficients. Compared with traditional global threshold algorithms, which can not extract the pavement crack exactly under non-uniform illumination, experimental results show that the denoising effect of this proposed method based on contourlet transform is better than that of other methods.2. Based on the mathematical morphology, multi-direction morphological structuring elements algorithm is proposed aiming to the different distributing characteristics between the noise and the disease pixel of the pavement image. The algorithm uses the method of the maximal between class distance to ascertain the varied gradient pixel. Then the erosion operation by using binary morphology is employed to obtain the edge and remove the noise. Compared with the traditional edge detection algorithm, proposed algorithm has restrained noise jamming effectively while detecting the road crack accurately and effect is fine.3. An image segmentation method based on multidirection morphological structuring elements algorithm is investigated to overcome the weak adaptability and the difficulty in selecting the adaptive threshold of the traditional image segmentation and to improve the reliability lacking in the traditional single evaluation of image segmentation, and the corresponding parameters are selected. Afterwards, segmentation evaluation method comprehensively considering various evaluation criteria is proposed. Experimental results show that the image segmentation method is of high accuracy and is suitable for the segmentation of varied images. Compared with the traditional single evaluation methods, the proposed composite method can evaluates the performances of segmentation algorithms more objectively and accurately.4. Based on different features of geometric shape, projection, crack distribution density and hole number are used to distinguish different crack. A pattern classifier based on RBF neural network is used to recognize different cracks according to the geometrical shape differences of different cracks. Compared with other image segmentation algorithms, experimental result indicates that the image edge is extracted, and noise is eliminated. Moreover, classification accuracy is achieved.5. Conventional detection method for crack parameters extract has many disadvantages. With the features extraction technology, defect area and parameters are worked out accurately. The experimental results indicate that defect area is extracted and noise is eliminated at the same time. Geometry parameter is accurately measured and an effectively means for crack defect quantification is provided.