Image-Based Crack Detection And Properties Retrieval For Railway Concrete Bridge

With the continuous growth of China’s railway mileage and the increase in service time of some rail lines, the inspection tasks of railway concrete bridge are becoming very heavy. Cracks on concrete bridge as a serious disease is an important part of the inspection. At present, the crack detection mainly relies on human, but this method has problems as low efficiency, low detection security, labor intensive, inaccurate results, and other issues. Perform crack detection using of image processing methods has been widespread concern in recent years. This paper presents a crack detection algorithm based on image processing, and realized the automatic information acquisition of the cracks, developed a testing equipment for crack detection of the underside of railway concrete bridge, which improves the efficiency of the crack detection for railway bridge. In this paper, the testing equipment designed by us is used for image capture of bridge bottom and camera calibration, then image processing techniques is used to analysis the captured image, to effectively detect the cracks in the image, and automated information acquisition is realized, the work of the crack detection for railway concrete bridges is completed. The device can effectively find out all cracks larger than0.15mm, satisfying the maximum limit of the bridge cracks.In the process of crack detection, the theory of liquid penetration has been used. After penetration depth study of the theoretical model, a new algorithm using percolation model is proposed, a binary image contains only crack information that we call cracks map is generated, by using morphological characteristics of cracks, the information of cracks can be kept mostly while the noise parts are ignored. Then, through distance transform and a series of morphological transformation processing applied to the crack map, combined with camera calibration results, the program can automatically determine the width of cracks.Finally, the receiver operating characteristic analysis (ROC) is performed to measure the performance of the crack detection algorithm, error analysis between the automatically measured data and manual measurement data is done to verify the effectiveness of the proposed algorithm.