Research On Damage Identification Method Of Concrete Bridge Based On Digital Image Processing

Crack is one of the most serious defects in concrete bridge structure.It not only affects the appearance of the structure,but also causes the corrosion of the internal reinforcement and accelerates the aging of the bridge,thus affecting the bearing capacity and safety of the structure.Therefore,rapid and accurate crack detection is an important means of concrete bridge detection and safety evaluation.However,the traditional crack detection technology is timeconsuming and laborious.With the development of science and technology,digital image processing technology has been widely used as a powerful tool for crack detection of concrete Bridges.In this paper,the original image recognition method is improved by in-depth learning of the basic theories of digital image processing and deep learning and applies to the field of crack detection,and explores the feasibility of digital image processing in infrared crack detection.Specific research emphases are as follows:1.Crack images were collected by compression bending experiments of concrete beams.The bending experiment of concrete beam was designed firstly,and the digital image of crack was taken by digital camera.Then,the crack information was identified by ACTIS system,and the effects of shooting distance,shooting Angle and focal length on the identification accuracy of the system were studied.2.A crack identification method of concrete surface based on block idea is proposed.First the image is divided into blocks,and then the method of average gray level difference between categories of segmentation is used to filter out sub-block,then two-dimensional maximum entropy threshold segmentation method is used to segment the image.The results can prove that the algorithm can accurately extract the crack target characteristics.3.Crack identification research based on infrared thermal image was carried out combined with digital image processing.Firstly,the infrared images of the crack were collected by an infrared thermal imager.Secondly,four infrared enhancement algorithms(grayscale linear transformation,histogram bidirectional equalization,infrared image enhancement based on gravity and suppression network,and infrared image enhancement algorithm based on adaptive histogram partition)were used to recognize the collected infrared crack images.It can be concluded that the improved infrared image enhancement method based on adaptive histogram partition has higher accuracy and can effectively locate the cracks and background in the image outside the distribution.4.An improved U-net neural network model is built based on deep learning theory.And pytorch database based on convolutional neural network theory,this paper constructs a kind of based on improved U-net neural network model,then the digital image of the crack obtained from the concrete compression bending test is flipped and marked,thus obtaining900 training set images and 100 test set images,then the model was used to train crack images.Finally,the superiority of the proposed algorithm can be proved by comparing the detection results with the results of original U-net and Segnet.5.The improved U-net neural network model is applied to the actual bridge detection.Firstly,the crack photos of Fengshancun Bridge and other Bridges were detected and collected in the field.Then,the improved U-Net neural network model was used to train and identify the crack images collected in the detection of Bridges.Finally,the crack width of the segmented experimental image was measured with Matlab programming.The results show that the average error between the crack width identified by the deep learning method in this paper and the actual results is 12.26%.