Research On Concrete Surface Crack Recognition Based On Machine Vision

Concrete,as a common and very important building material,has many advantages,such as good durability,easy processing,and low cost.It is widely used in various infrastructure construction such as dams,highways,and bridges.However,with the continuous increase in the service life of concrete components,the degradation rate of concrete components such as bridges,roads,and dams is accelerating due to factors such as environment and loads.Cracks,as a common safety hazard affecting the structural quality of concrete components,do not pose a significant threat to concrete components themselves.However,without timely and effective treatment,concrete cracks are likely to cause other more serious disasters.Once concrete components such as bridges and dams collapse,it will cause serious safety accidents.Therefore,timely monitoring of cracks is of great significance for ensuring the safety and stability of concrete components.The traditional method of manually detecting cracks is time-consuming and labor-intensive,and the results are also highly subjective.However,non-destructive testing equipment is expensive,and the testing process is greatly affected by external conditions.The use of image processing technology for detection has the advantages of low cost,high flexibility,and high detection accuracy,making it possible to widely apply detection methods based on image processing technology.However,generally speaking,an image detection algorithm can only identify cracks in specific situations,and its universality is not high.Therefore,this article conducts in-depth research on algorithms for concrete crack recognition,mainly including:(1)Various pre-processing algorithms of concrete crack image are studied.By comparing the filtering effects of median filter,Gaussian filter,Bilateral filter and mean filter algorithm on concrete crack image,Gaussian filter algorithm is finally selected to smooth the gray image of concrete crack and reduce image noise.By comparing the enhancement effects of different image enhancement algorithms on concrete crack image,it is found that,the gamma transform algorithm can better improve the contrast of crack images and facilitate subsequent crack segmentation and recognition.(2)Research on concrete crack segmentation and extraction algorithms has been carried out,and two improved fuzzy clustering algorithms have been implemented to segment concrete crack images.The segmentation performance of edge detection algorithms,threshold segmentation algorithms,and fuzzy clustering algorithms for concrete crack images has been compared and studied.The results show that the fuzzy clustering algorithm produces less noise in the segmented binary image and has better segmentation performance.Then,combining the area size and circularity of the connected domain,the binary image is filtered and denoised to accurately extract concrete cracks.The experimental results show that the Dice coefficient of the HMRF-FCM algorithm can reach an average of 0.9017.Finally,the density,center of gravity,and other parameters of concrete cracks were calculated.(3)Finally,in the Python development environment,this article uses Python and Matlab to jointly program,and combines OpenCV library,PyQt,and MySQL database to complete the development of concrete crack identification software.The software encapsulates OTSU algorithm,FCM algorithm,FLICM algorithm,and HMRF-FCM algorithm,which can extract concrete cracks using different segmentation algorithms according to different conditions.After experimental verification,concrete cracks can be accurately identified and extracted.The software automatically stores the obtained concrete crack density,center of gravity,and other parameters in the MySQL database,and the final calculation results are displayed in the table on the interface.