Abstract: Traditional bridge crack detection methods are of high cost and high risk. A bridge crack detection and classification method was proposed based on a climbing robot using image analysis with a miniature camera mounted on the robot to collect images. First, the motion blur of acquired images was removed by Wiener filtering method. Second, wavelet transform was used to enhance the fractures of the crack in the image. Third, to complete crack image recognition, the surface morphology analysis is applied to extract crack fragments and then KD-tree was used to connect them. Finally, support vector machine method was used to classify crack images based on a series of basic visual characteristics and geometric features. Comparison of geometrical characteristic classification method and BP neural network classification method, results show that our method is better.

Key words: climbing robot, motion blur, wavelet transfourm, surface morphology analysis, KD-tree, SVM