Research On Image-based Recognition Method For Switches On Protecting Panels In Substations

Electricity is an important energy source for people's livelihood.The generation,transmission and the utilization of electricity play crucial roles in maintaining social stability.Under the guideline of “Strong Smart Grid”,the maintenance and monitoring of substations and power systems are becoming more and more intelligent.During the processing of making substation more intelligent,a large number of existing equipment can be upgraded automatically by applying machine vision technology,avoiding duplication of investment,and the intelligent level of substation monitoring system can be improved at a lower cost.This paper takes the image of the switch on protecting panel as the object of the research,the whole process of automatic recognition of the switch state is realized by using the related methods of image processing,which contains correction,location and segmentation of a single switch area,and state recognition.The main research contents are the following aspects:(1)The research on correction method on the protecting panel images.Because of the inconsistency of viewing angle,illumination or other factor,the correction of the captured images is need to utilize,which contains illumination correction and viewangle correction.And after correction,the image is mainly containing switches with front-view angle,so that the segmentation of the switch area and the recognition of the switch state can be realized.Based on this,some correction methods on images are studied in this paper,especially the method on angle correction.And two methods for correcting the angle of protecting panel images are proposed,and experimental is made to verify the correction effect.(2)The research on the methods on location and segmentation of switches on protecting panel.The edge feature,texture feature,shape feature of corrected images,along with the feature extraction method are studied.The regularities of distribution of extracted features of switches are also studied.Based on the studies,three methods on location and segmentation utilizing different features are proposed: one for square switch with edge feature,one for circle switch with mark labels,and the last one for both type of switches with pattern matching.Experiments are conducted in order to verify the effect of each method.(3)The research on the methods on state recognition of switches.After the location and segmentation,the statistical models of relevant features are established for the image of a single switch area to determine classification criteria.Some methods on feature classification are studied.Then the method for state recognition is proposed.Finally,the optimal feature description and classification methods are selected through comparative experiments.And experiments show that the proposed method can extract features and recognize states of a single switch image effective and accurately.