Research On Bridge Crack Recognition And Measurement Method Based On Image Processing

Cracks are one of the main diseases of concrete bridge,and the detection of bridge cracks is an important link to evaluate whether the bridge can serve safely.The methods of manual detection and bridge inspection vehicle detection have disadvantages such as low efficiency and influence on road traffic.With the rapid development of computer,image processing technology has become more and more widely used.In this dissertation,an in-depth study of bridge crack detection methods is carried out,and according to the characteristics of bridge cracks,a method for identifying bridge crack energy based on image processing is determined,and it has been effectively verified.The main work of this dissertation is as follows:Firstly,the measurement principle of image processing is studied and the lens distortion model is established,and the camera calibration and size calibration are realized.The principle of preprocessing and the characteristics of noise shape are studied,and the crack extraction method in this dissertation is determined.According to the characteristics of different types of cracks,the methods of image post-processing are analyzed.The calculation indexes of the geometric parameters of the cracks are clarified,and the crack image classification and measurement methods are established.By studying the grayscale processing and Mask homogenization algorithm,a preprocessing method based on the improved Mask algorithm and slicing processing algorithm is determined,which achieves the image illumination homogenization,enhances the contrast between cracks and background,and retains more details of the cracks.The preprocessed binary image contains multiple types of noise.This dissertation effectively filters out various types of noise based on the feature combination method and the improved connected domain method.Through the image fusion algorithm and the crack stitching algorithm,the connection of the broken crack fragments is realized,and the image containing only complete cracks is obtained.Finally,through in-depth analysis of a large number of extracted fracture images,various fracture features are summarized,and SVM decision tree algorithm based on HOG feature and geometric feature fusion is proposed to achieve the task of fracture image classification.The image post-processing is realized by Hilditch thinning algorithm and burr removal algorithm,and a smooth skeleton image is obtained.Combining the characteristics of the crack direction and the idea of substituting straight for the curve,a measurement algorithm based on skeletonized images is proposed,and the maximum width,length and area ratio of the crack are measured.Based on the research content of this dissertation,a crack image recognition system is designed and visualized operation is realized.Through comparative experiments,the results show that the bridge crack identification method based on image processing in this dissertation can effectively complete the crack identification task,and the obtained image classification accuracy and parameter calculation results meet the national standards and requirements.The method in this dissertation greatly improves the detection efficiency and accuracy,reduces the detection cost,and ensures the safety of the inspectors.