Research On Key Techniques Of Components Detection Algorithm In Power Line Images

Power line inspection is crucial to troubleshoot possible faults and safe hidden trouble in the transmission lines,it can discover the defects existing in the transmission line as early as possible,and avoid the happening of serious electrical accidents.With the rapid development of science and technology,electric power company can make use of some high-tech to solve the problem that the traditional manual inspection transmission line is inefficiency,and improve the detection accuracy.It can quickly and accurately to gather the video and image data of transmission lines by helicopters or unmanned aerial vehicle(uav).After that,to process these massive aerial power line inspection images by computer vision technology.According to the different goals,transmission line inspection images can be divided into power line line and insulator image.The background environment of aerial power line inspection image is very complex,and complex background will be a great deal of interference to detect target in the image,so the traditional target detection algorithm in dealing with this kind of aerial power line inspection image is difficult to obtain satisfactory results.In order to solve these problems,through research and analysis a large amount of power line images,and combined with the methods of some related literature,we put forward the following several kinds of innovative algorithms in different goals and different point of view.The main contributions in this thesis can be summarized as follows:1.The fault diagnosis of power lines and related accessories can be divided into power line detection,related accessories detection and all kinds of fault diagnosis.This paper proposes an integration algorithm that includes all kinds of detection methods and fault diagnosis methods.We use a line segment detection method Based on the voting strategy to detect power line.Then,we find out the suspected fault location of the line by gray statistical characteristic.Finally,to realize fault diagnosis using spatial relations characteristics and haar-like features.The experimental results show that the algorithm has the ability of fast accurate detection and fault diagnosis,and is also applicable to a variety of target detection,and a variety of fault diagnosis.2.Local features widely used in target detection,also used in the insulator detecting.First,we propose a improved insulator segmentation algorithm based on local features.Although based on active contour models(ACM)segmentation algorithm has good segmentation of insulator in power,but there is good performance of the algorithm in processing large resolution insulator image with complex background.To realize coarse segmentation by using local feature matching,then achieve fine segmentation by using insulator image segmentation algorithm based on active contour model processing.It can not only improve the processing speed of the original algorithm,and has the ability to handle large resolution insulator image with complex background.Besides,we propose a robust insulator detection algorithm based on local features and spatial orders for aerial images.First,we detect local features and introduce a multi-scale and multi-feature(MSMF)descriptor to represent the local features.Then,we get several spatial orders features(SOF)by training these local features,it improves the robustness of the algorithm.Finally,through a coarse-to-fine matching strategy,we eliminate background noise and determine the region of insulators.Finally,the experimental results show that this algorithm has more anti-interference ability and higher accuracy compared with other insulators detection algorithm with different background environment.3.Insulator detection algorithms are mostly based on gray image,however,through the study of a large number of insulator image analysis,we can extracted more features from the color information of insulator.Therefore,we propose a simple and effective segmentation algorithm based color clustering and texture features to solve this problem.In the first phase of the segmentation,the insulator image is segmented into several parts using K-means clustering and the color features derived from RGB,HSI,and L*a*b*color spaces.In order to facilitate the following work,each part can be divided into several disconnected areas.In the second phase,we extract texture features of insulator using GLCM and bag-of-words(BOW)model.The GLCM is used to locate insulator areas,and the BOW model is combined with a fast sliding window strategy and support vector machine(SVM)for the accurate segmentation of insulators.Experimental results show good capabilities of the proposed algorithm in segmentation of aerial insulator images,and indicate the possible use of our method in practical applications.4.Detecting the defect of insulator can be divided into the insulator detection and insulator defect detection.First,we can detect the regions of insulator in input image by the proposed insulator detection algorithm.Then,extract local feature points of insulator,fitting a straight line with all feature points of the insulator,and obtain angle of the insulator.Finally,we can find the part of the defect by the result of projection transform of the insulator.The experimental results show that the algorithm can accurately remove the noise of background,and detect the damage of insulator.