Research On Apparent Crack Detection Method Of Bridge Based On Image Fusion

Apparent cracks in bridges are one of the common diseases of bridge structures and can seriously affect the safety and reliability of bridge structures,so crack detection for bridges is becoming increasingly important.With the development of computer technology,crack detection technology is also constantly improving,making up for the shortcomings of manual detection.In this paper,using computer technology,the image preprocessing method is first studied to improve the image feature information.Then,the CNN deep learning network is used to complete the crack recognition and classification.In order to better extract the crack skeleton,combined with the image pixel-level fusion algorithm,the identified crack image is pixel-level fusion to enrich the features of the identified crack image.After pixel-level fusion,image segmentation is used to extract crack images more accurately.Finally,the quantitative study of the crack image after segmentation is carried out,and the pixel length and width of the crack image are obtained.Based on the above research,an automatic segmentation method for crack images is proposed to obtain the length and maximum width of multi-segment crack images.In this paper,the crack image processing and extraction are explored in detail,aiming to achieve accurate extraction of cracks in concrete bridges and provide more ideas for crack image research.The main research content includes the following five aspects:(1)Research on crack image pretreatment method of concrete bridge.Grayscale processing,image filtering processing,image enhancement methods,etc.are used to obtain more and clearer image detail information.(2)The artificial intelligence algorithm is used to realize the image classification and recognition of concrete cracks in bridges.The convolutional neural network model is used to train and verify the crack image dataset,and the network model combined with multi-scale feature fusion is combined to improve the accuracy of crack recognition and classification,and the accuracy of the final result is higher.(3)Research on pixel-level image fusion methods.The pixel-level image fusion method is used to improve image details,and a variety of pixel-level image fusion methods are comparatively studied,including non-multi-scale and multi-scale methods.Obtain image information that has undergone pixel-level image fusion.(4)Bridge concrete crack image segmentation.A variety of image segmentation methods are used for the crack image after pixel-level image fusion,and a segmentation method suitable for the bridge crack image is found.It mainly includes four types of concrete bridge crack segmentation methods(edge,threshold,area,and bionic algorithm),and the effect of each image segmentation method is obtained.(5)Quantitative study of crack characteristics.Mainly for a series of previous studies on crack images,the excess non-crack area is removed,a relatively complete crack area is retained,and the maximum inscribed circle algorithm is used to obtain the maximum pixel width of the crack.Then,the skeleton extraction method is used to refine the binary image to obtain the skeleton map of the crack target,and the pixel length of the crack is obtained by counting the number of pixels in the skeleton map.In addition,on this basis,this paper proposes to measure the maximum width and length of cracks in segments,and obtain the maximum width and length of cracks in the required segment by automatically segmenting the crack image.