Research On Extraction And Recognition Method Of Damage Feature Of Pipeline Inner Wall

Water supply pipelines are an important infrastructure for the development of modern cities,and in order to ensure its safe operation,it is of great industrial application value to achieve accurate detection of damage types on the inner wall of pipelines.Nowadays,image processing and pattern recognition technology can effectively detect the types of damage in the inner wall of the pipeline,but there are problems such as low feature extraction accuracy,poor recognition rate,and subjective factors.To solve the above problems,in this paper,an improved random forest feature selection algorithm and support vector machine classification model is proposed to achieve feature extraction and automatic identification of damage images of water supply pipelines and provide important technical support for pipeline maintenance.On the basis of analyzing the structural characteristics of the water supply pipeline,firstly,the endoscopic pipe image acquisition platform is built,and then the panoramic expansion algorithm based on the cone bidirectional projection model and the Retinex algorithm based on the dual illuminance estimation model are used to preprocess the image,thereby improving the contrast of the image,and finally the image sample library of the damaged inner wall of the pipeline is established to as identification and classification samples.The random forest algorithm has a low recognition rate when selecting the damage features in the inner wall of the pipeline,in this paper,a random forest algorithm based on feature simplification is proposed to calculate the feature weights,evaluate and sort the importance of the extracted statistical properties of color channels,gray-level co-occurrence matrix,and graylevel run lengths,and selects the features according to the importance sorting results.This algorithm improves the credibility of attribute weights when dividing more feature data and realizes feature dimensionality reduction.The results show that the improved random forest algorithm can reduce the redundant features of the image and effectively improve the recognition accuracy of the damage features of the inner wall of the pipeline.The support vector machine classification model has poor performance in the identification of the damage characteristics of the inner wall of the pipeline,in this paper,a support vector machine classification model based on the slime mold optimization algorithm(SMA-SVM)is proposed,which optimizes the penalty parameters and kernel function parameters of the support vector machine and improves the generalization performance of the SVM.Experimental results exhibit that the accuracy of the SMA-SVM classification model is as high as 94.710%.Compared with the SVM classification model,the SVM classification model based on differential flower pollination optimization,the SVM classification model based on particle swarm optimization,and the BP neural network,the recognition accuracy of the proposed algorithm is improved by 4.786%,3.023%,4.030%,and 0.503%,respectively.We can draw the conclusion that the support vector machine classification model based on the slime mold optimization algorithm has a good recognition effect on the damage image of the inner wall of the pipeline.