Research And Design Of Pavement Crack Detection System Based On Image Processing

With the rapid development of Chinese economic construction, road construction in the national economy plays a more and more important role. The road maintenance and management become more prominent, and the detection of pavement damage become one of the important work to the road maintenance department. In addition, as of a number of early constructed highway into the period of repair or overhaul, in order to improve the service life of the road, getting the damaged road surface data is becoming a most important work to road maintenance department. At present, pavement Distress detection method in our country is still artificial detection, but the method of artificial detection of the presence of low detection efficiency, affecting the normal traffic and the intensity of work, time consuming, unsafe and so on. Therefore, the study on the damage of pavement detection system has the certain practical significance. This paper mainly introduces the research work of pavement crack detection based on image processing.This paper first introduces the research status and development trend of the detection system for pavement damage, and introduces the related specification type asphalt and evaluation standard of our country at present. This paper focuses on the part of the pavement damage detection, i.e. pavement crack detection. By reference to domestic and foreign related testing equipment, we put forward the crack detection method based on machine vision in the hardware system. And the system mainly includes the following parts:pavement image acquisition system, GPS geographic information system, computer image processing system and lighting system.This paper studies several digital image processing algorithms, including image enhancement, the image edge detection as well as the crack image segmentation. Digital image processing is the focus of this research, but also the core of the pavement crack detection. In this paper, we analysis and compare the advantages and disadvantages of the different operators. Image enhancement processing can effectively reduce the noise in the image, and the image processed is in favor of image segmentation and recognition of the following. In image segmentation, several segmentation methods are introduced and compared, and we choose the Iterative threshold segmentation method.Finally, we research on the classification of pavement crack image, and introduce several pattern recognition methods including:statistical pattern recognition, fuzzy pattern recognition, neural network pattern recognition and pattern recognition method of support vector machine. This paper chooses the BP neural network as the classifier to classify the pavement crack image. We designed a three layers BP neural network, and provided three kinds of features extracted as the input layer, and the four output including:transverse crack, longitudinal crack, reticular cracks and no cracks. By the simulation of MATLAB, the results showed that the BP neural network classifier can effectively classify the pavement crack classification and high accuracy.