Research On Key Technology Of Concrete Crack Detection Based On Image Processing

As time goes on,the concrete materials of bridges are aging,and then cracks and other diseases appear,which constitute the safety risks of bridges,and even lead to the collapse of bridges,resulting in economic losses,casualties and social unrest.In recent years,convenient bridge crack image detection system and method has become a research hotspot.However,the carrying and installation of the bridge crack detection system is still cumbersome,and the crack detection algorithm still can not meet the needs of the actual scene.The complex background at the bottom of the bridge brings a great test to the anti-interference ability of the detection algorithm.Aiming at the problems of poor reliability and real-time of existing bridge crack detection algorithms,this paper proposes three real-time detection algorithms of highresolution crack image based on embedded platform.Firstly,Gaussian filter and moving average method are used in the image preprocessing to denoise and segment the crack image.Then,the geometric features of the connected components are described according to the geometric moments,and the background contour interference is filtered by combining the region growing method with geometric moments.Cracks under complex background are segmented into discontinuous fragments.Aiming at the problem that the crack fragments are similar to the background spots,a spatial aggregation model(SAM)is proposed based on the spatial adjacent features and the similar extension trend of crack fragments.According to the spatial similarity,the adjacent candidate crack fragments are aggregated recursively.The experimental result shows that the SAM algorithm is significantly ahead of several existing algorithms,and its F-Measure is increased by more than 107%.Dealing with an image with 15 megapixel on the embedded platform,it costs only 1.51 s.Under the exploration of SAM algorithm,it is found that the screening and filtering process of crack fragments is actually equivalent to a clustering process,and the crack fragments belonging to the same crack are clustered into a cluster.According to the spatial distribution characteristics of crack fragments,the spatial cost index and relaxation index are defined,which constitute the distance function of crack fragments.A real-time detection algorithm HCSR(hierarchical clustering using spatial cost and correlation ratio)based on hierarchical clustering is proposed to cluster crack fragments.The experimental result shows that HCSR algorithm has the same detection performance as SAM algorithm,while the running time is reduced by 46.8%.The cost time of processing a single image is reduced to0.803 s,which effectively improves the real-time performance of bridge detection algorithm.Based on the SAM algorithm and the HCSR algorithm,dynamic fusion clustering using hierarchies model(DFCH)is proposed.Firstly,the crack fragment model and crack cluster model are defined.Then,the cluster fusion operation in the clustering process is redefined.The distance between clusters is calculated by using the overall attributes of clusters.Compared with HCSR algorithm,The accuracy of DFCH algorithm is improved by 17.7%while keeping the same recall rate.Compared with SAM algorithm,DFCH algorithm improves the recall rate by 23.0% while maintaining the same accuracy,and basically inherits the speed of HCSR algorithm.The cost time of processing a single image is 0.849 s.DFCH algorithm effectively combines the advantages of SAM algorithm and HCSR algorithm.The proposed SAM,HCSR and DFCH algorithm significantly improves the reliability and real-time performance of the bridge crack detection algorithm.They provide some idea for real-time detection of bridge cracks based on embedded mobile platform.