Bridge Disease Image Processing And Identification

With the rapid development of national infrastructure construction,the construction of concrete bridges has presented blowout phenomena and the bridge structure has become more complex.Due to the high-intensity use,the maintenance of bridge facilities has become an increasingly important link.How to quickly and efficiently find bridge diseases,and then develop a corresponding repair strategy for different types of diseases,and then bridge maintenance becomes a bridge maintenance department faces Big problem.The traditional method of detecting diseases based on artificial vision is far from meeting the needs of huge workloads,and it also consumes material and financial resources.Therefore,the use of digital visualization of disease detection methods to complete the assessment of the degree of damage to the bridge is a new trend.In view of the complexity and diversification of bridge disease images,digital image processing methods are still a research difficulty.This thesis mainly deals with the image after obtaining the bridge disease image,so as to classify and identify the disease types.The research contents are as follows:(1)First,the acquired disease images are preprocessed,including the use of weighted average method for graying,gray histogram technique for grayscale correction,and median image filtering method for modified images for denoising and smoothing,etc.The operation,through these pre-processing methods,highlights the disease information in the image.(2)After the preprocessed image is obtained,the image is divided.Analyze several commonly used segmentation methods,and propose a segmentation algorithm based on adaptive thresholding when the segmentation results are not ideal.The segmentation effect is significantly improved and the disease target can be segmented relatively completely.(3)Feature extraction is performed on the segmented image object.According to the texture features of the disease,the gray-level co-occurrence matrix method can be used to obtain the texture feature information such as energy,entropy,moment of inertia,and correlation of the disease,and the extracted feature information is used to build a feature database.(4)Introduce several commonly used methods of classification and identification.After comprehensive comparison,the support vector machine method(SVM)is introduced.The linear separability and linear inseparability of the SVM method are deduced in detail.The three kernel functions that solve the nonlinear separable sample problem are derived.Finally,using Lib SVM software package containing three kinds of kernel functions combined with "one against one" multi-classification decision can better classify the data information in the signature database.Through the above methods of classification and identification,experiments are performed on the MATLAB simulation platform.The results show that the SVM based on radial basis kernel function can obtain good classification recognition effect and can accurately identify bridge disease information.