Research On Concrete Bridge Crack Detecting Algorithms

With the increase in the number of bridges and the passage of time,cracking has become a common problem in concrete structure bridges.If left unchecked,it will lead to significant accident risks and economic losses.Bridge crack research has become particularly important.Due to the influence of factors such as bridge material and environment on crack measurement,existing research methods for bridge cracks are not ideal in dealing with noise and burrs,Therefore,efficient and accurate crack detection has become a crucial research topic.This article focuses on improving algorithms for crack segmentation and edge detection in bridge crack images.First,the concrete bridge crack image is preprocessed to realize the graying and histogram equalization of the crack image,eliminate the influence of non crack factors such as shadows,and the bridge crack image is processed by filtering and image enhancement,laying the foundation for further image processing.Second,a U-shaped network based on residual neural network Res Net101(URes101)was employed for bridge crack segmentation after examining the structures and properties of fully convolutional neural networks(FCN)and U-shaped networks(U-Net).The U-shaped residual convolutional neural network was given a parallel residual structure,and a better version of the URes101 crack segmentation method(PURes101)was suggested.The extracted features at different scales and up sampled recovered features were fused.The simulation results showed that,the PURes101 network enables better segmentation of small and weak edges in images,with an accuracy improvement from 94.5% of the URes101 network to 96.1%.Applying Sobel,Canny,and crack center point method(CCPM)to extract crack edges,an improved weighted fuzzy entropy edge detection method is proposed to address the problem of heavy burrs in CCPM.Several different information measures are proposed based on the nature of crack edges.Through the principle of block splicing and CCPM’s crack feature determination method,the original algorithm threshold and Gaussian convolution threshold are weighted and calculated,the improved threshold was applied to detect and concatenate the sub block images one by one.The simulation results verified that the improved weighted fuzzy entropy crack edge detection method can better refine and smooth the edges while completing crack edge recognition and extraction,reducing the impact of independent noise and burrs.Finally,the projection method was applied to classify cracks in concrete bridges.After performing morphological closed operations to eliminate gaps and holes,and implementing the Hilditch refinement algorithm to refine cracks,crack information was extracted,and a GUI user interface design was completed to help engineers use it more conveniently and efficiently.