Research On Contamination Detection Technology Of Insulator Based On Image Processing

Especially in China,the climate is changeable,the terrain is complex,and the pollution in some areas is serious,which leads to the contamination problem of insulators and it is particularly serious.The pollution on its surface make it is easy to take place pollution flash,which leads to a large area of power outages,and then people's production and life will have an immeasurable result.Therefore,it is necessary to detect and identify the contamination of insulators.In this paper,image recognition,precise location segmentation,information collection,classification of insulator pollution level and pollution level recognition are finished by various technical means,which realize the high efficiency and high intelligence of insulator pollution detectionFirstly,the insulator image information is collected by a camera.The image binarization,filtering the noise,and other pre-processing processes are used to reduce the environmental interference by a computer.The insulator in the image is identified by the SURF image matching technology,and preparation work is completed.Secondly,as the commonly used image segmentation method for insulators,the maximum threshold difference method and the k-means clustering algorithm have some problems that can't remove the complex information of insulator images,the large error segmentation rate,and the certain requirements for the background of the image measurement exist.A method to improve the accurate positioning and segmentation of insulator images based on the CV model is proposed to achieve insulator image segmentation.Experiments show that this method has certain advantages,which can solve the problem of separating insulator images by using the maximum threshold difference method and k-means clustering algorithm,and obtain the target image.By counting the RGB color component and HSI color component of the insulator target image,the characteristics of the insulator image are obtained.And then sample eigenvalues are reduced after information fusion using the Principal Component Analysis method.FCM clustering is performed to choose and obtain standard sample data.Finally,the application of multi-type SVM classifier one-to-one voting strategy is proposed to train the standard sample data to complete the classification of insulator pollution level.And the establishment of the classification detection model of insulator pollution level is completed.Insulator pollution level can be carried out through this model.The existing red-brown bar insulators in the laboratory is made use of carrying out artificial smear experiments to obtain sample images.The insulator pollution level detection system is designed based on MATLAB R 2014 a GUI as a platform through theoretical analysis,Which can realize the automation of pollution classification of insulators.And this system can correctly realize the identification of pollution level,and improve the efficiency of pollution detection of insulators.