Study Of Pavement Crack Detection And Identification Based On Image Segmentation

Pavement crack detection plays a vital part in expressway maintenance andmanagement. Thanks to rapid development of computer and video detection technology,the real-time sampling and recognition of pavement crack has become possible.However, content of pavement crack image is so complex, and there is also much noise,the traditional image processing method is impossible to slove the problem properly. So,image processing algorithm is also the priority research content in pavement crackimage detection.As the direction of crack in image is so complex to recognize, and it always hasmuch noise, gray crack image detection algorithm is the most important and difficultpoint. Currently, most proposed processing algorithms are complicated to implement,and their processing efficiency is not so obvious. To achieve the balance betweenprocessing efficiency and complexity, this paper conducted a series of in-depthresearch，and its main work are as follows：Based on image segmentation, it made researches on methods of pavement crackdetection，including improved median filtering algorithm, pavement crack imagesegmentation method based on the average of Multi threshold, and multi-directionstructuring elements based morphological image restoration method. Results show thatthe proposed detection method can effectively remove the isolated noise point, smooththe edge and improve the segmentation accuracy, with the crack shape, length and widthunchanged. The method provids a solid foundation for the further feature analysis ofroad pavement crack image.By the help of binary image projection analysis principle，it could accuratelyrecognize the type of the processed cracks，and measure the feature of segmented crackimage, including length and width of the regular crack， and area calculation ofirregular crack.Finally, to test total effect of the proposed crack image processing method，itdesigned the experiment. Take longitudinal and block type pavement crack for example,it used the method for crack measurement, processing and recognition. Results showthat, the method is not only simple but also effective, it could commendably satisfy therequirements for practical road crack detection，and make completely and accuratelyidentification for the regular crack image. But for the irregular road surface crack， it also needs further research and experiment to improve the accuracy of their deeper levelclassification and recognition.