Abstract:Aiming at the problems such as weak anti-interference ability of traditional crack detection algorithm and easy misjudgment of shallow cracks, a local feature clustering combined with area growth algorithm for bridge crack detection is proposed. Firstly, in order to solve the problem of crack segmentation affected by the concrete cuticle shedding and water seepage, the Gauss-Frangi dual filter was used to blur the image, degrade the characteristic information of noise, and enhance the linear structure in the image. Secondly, aiming at the problem of shallow cracks with weak features that cannot be recognized by conventional algorithms, according to the spatial correlation between crack points in local regions, a new algorithm based on grid clustering combined with regional growth algorithm was proposed to realize the dynamic segmentation of local regional cracks. Finally, a method based on shape feature and structure similarity principle is proposed to eliminate the pseudo-crack noises in the image segmentation. Experiments show that the proposed algorithm can detect more details of cracks, maintain a high accuracy rate, and improve the quality of crack image segmentation.