Research On Insulators Target Extraction Technology Based On Infrared Image

With the development of the power industry,insulators have been widely used in power systems.In the power transmission system,the existence of degraded insulators directly threatens the safe operation of the power system.The failure of the insulators will cause huge economic losses and even human casualties in the power transmission system.Regular inspection of insulator status is a necessary indemnification for power system operation and maintenance.Traditional detection methods,which limited by poles and towers,are susceptible to electromagnetic interference and need to be powered off.In recent years,infrared detection technology for fault detection of power equipment has attracted widespread attention.As a non-invasive method,infrared imaging method has the advantages of visualization,safety,non-contact,and interference-free,which is convenient for monitoring equipment operating status and on-site detection.In order to detect the defects of insulators,such as cracks,surface contamination and damage,identifying and locating the insulators from the images collected on site is a key link in fault detection.Thus,research on the insulator target extraction algorithm based on infrared images can lay a foundation for insulator fault detection.The infrared images of the insulators in the power transmission system has high noise and low contrast.In order to improve the accuracy of the insulator extraction in the power transmission system,the main research contents of this thesis are as follows:(1)A classification model of infrared image complexity based on BP(Back Propagation,BP)neural network is implement.The collected infrared images are classified according to the image feature parameters through the BP network,which solves the problems of long time and code waste caused by the same algorithm for all images,and provides a strong guarantee for subsequent target extraction.(2)An insulator target extraction method based on improved FCM(Fuzzy Cmeans,FCM)algorithm is proposed.It solves the problem of effectively extracting target insulators in a simple background,which usually contains independent insulators and does not cross the tower.The improved FCM algorithm is used to cluster the SURF(Speeded-Up Robust Features,SURF)feature points of the poles and insulators,according to the shape feature values of the insulators,the insulators are identified and located,and finally the accurate extraction of the insulators is realized.(3)An insulator extraction method based on improved template matching is proposed.It solves the problem of effectively extracting target insulators in a complex background,which usually contains multiple insulators of different sizes and different angles.An improved image descriptor is proposed.The insulator template image library and the test image library are matched to achieve accurate extraction of insulators through the matching strategy from coarse to fine.Through qualitative analysis,the insulator extraction results by this method are compared with the experimental results by the traditional clustering algorithm and template matching algorithm.The experimental results prove that the extraction accuracy of this algorithm is higher.Through quantitative analysis,three indicators of true positive rate,false positive rate and accuracy rate are calculated,which further confirms the validity and accuracy of the method in this thesis.