The Study Of Electric Power Equipment Image Feature Extraction And Classification Method

With the rapid development of intelligent electric grid construction and anincreasing number of types of electric power equipment, feature extraction andaccurate classification of electric power equipment, help to improve the automationand intelligent level of power system management.This paper first preprocess to the collected electrical equipment image,including normalizing image size, enhancing image contrast, filtering noise andother processes, in order to enhance useful information and get a higher resolution,contrast, stronger, more prominent image objects.This paper then study the color、texture and shape feature extraction of image.And extract features include color moments based on the HSV color space, texturefeatures based on statistical co-occurrence matrix and Hu invariant moments.Aiming at the problems that applying Hu invariant moments to extract imagefeature computes large and applying corner vectors to match has too dimensions,this paper presented Harris corner invariant moments algorithm. This algorithmonly calculates corner coordinates other than the entire image coordinates, so canchange the point feature into feature vectors, and reduce the corner matchingdimensions.Then through the research of the relevant principles of support vectormachine, constructed a SVM classifier and selected RBF(radial basis function) askernel function. Use the default parameters, Grid Search method and ParticleSwarm Optimization Algorithm to optimize the parameters of SVM kernel, andcompare their capability in kernel parameter optimization. We conducted anumber of classification experiments for single and combined features, and finallychose color moments and corner invariant moments as integrated extractionfeatures. The result shows that combined with SVM optimal parameters byParticle Swarm Optimization, can achieve better classification accuracy, and theaverage classification accuracy rate can reach more than85%.