Research On Defect Detection Algorithms For Patrol Inspection Image Of High Voltage Transmission Lines

In order to ensure the continuity,safety and reliability of power supply,it is very important to locate malfunctions in high voltage transmission lines.Most Current maintenance work on high voltage transmission lines still need manual checking.Manual checking malfunctions has some problems,such as low work efficiency,low detection accuracy,and serious resource consumption.Therefore,this paper applies the knowledge of image processing and deep learning to the malfunctions detection of high voltage transmission line.Three novel methods of bird's nest and vibration damper detection on the high voltage transmission line are proposed.The main work is as follows:(1)Aiming at the problem of automatic detection of bird's nest on the high voltage transmission line,a novel method of bird's nest detection based on cascade classifier and combination features is proposed.By analyzing the different features of the bird's nest and iron tower,the following four novel features are proposed:proportion of total white area(PWA),ratio of white pixels(RWP)in each lap,projection feature(PF)and improved burr feature(IBF).These features are used to describe the characteristics of the bird's nest backbone area and the edges,respectively.In addition,the cascade classifier combined with four proposed features is used for further classification of bird's nest region.The proposed detection method consists of three stages:first,the suspected bird's nest region is obtained by template convolution.Second,PWA and RWP features with low dimensionality and high discrimination are used to classify the sample set of suspected bird's nest region.Third,PF and IBF features are adopted for secondary classification to the previous classification results to reduce the misclassified samples.Experimental results show that the proposed method can accurately detect the bird's nest and achieve good performance.(2)Aiming at the automatic detection of the vibration damper of high voltage transmission lines,a cascade classification algorithm based on the combined features of block-based Haar(BHaar)and region-based local binary pattern(RLBP)is proposed.In view of the characteristics of the vibration damper on the high voltage transmission line,BHaar feature and RLBP feature are proposed.These features are combined with Histogram of Oriented Gradient(HOG)features to detect vibration damper.The algorithm is as follows:first,preprocessing operations(including denoising,brightness adjustment,and size adjustment)are performed on the unmanned aerial vehicle image.Second,the improved normalized cross correlation matching algorithm is used for template matching.Scale and rotate the template during the matching process to obtain the suspected vibration damper region.Third,HOG feature is used to classify the sample set of suspected vibration damper region.And the BHaar feature and RLBP feature are used to do the secondary classification to the positive and negative sample.Fourth,statistics of the final classification results.Experimental results show that the algorithm can detect the vibration damper on high voltage transmission lines efficiently and accurately.(3)Aiming at the problems of difficult classification feature design and inefficiency in traditional methods of target detection,a novel method of vibration damper detection based on deep learning is proposed.We chose the YOLOv3 network model as the learning framework of vibration damper detection.In order to improve the effectiveness of the algorithm,preprocessing operations(including noise removal,brightness adjustment,and direction correction)are performed on the test image.Experimental results show that the vibration damper detection algorithm with increased preprocessing operation can effectively reduce the missed detection rate and has higher detection accuracy.