Research On Algorithms Of Vibration Damper Detection On Power Line Image

In the smart grid system,the intelligent inspection of power lines is an extremely important part.The lightning hammer on the power line is an important part of the equipment to ensure the normal operation of the power supply in order to reduce the installation of the power power line due to the vibration of the power power line.Therefore,the research on the detection of damper on the power line image shows the important value.The traditional power line inspection work is basically a manual inspection way,this mode of operation labor intensity,high risk factor and low efficiency.In recent years,UAV aerial patrol way by some power companies to use.However,UAV images are vast,background complex,viewing angle,target resolution is not the same,which gave the shock hammer detection has brought great challenges.Based on this background,this paper studies the relevant work at home and abroad,and studies the detection and positioning algorithm of shock hammer on power line to solve the problem of robust detection of vibration hammer in complex background,prepare for defect diagnosis of damper.The main work of this paper is:(1)Detection of damper based on AdaBoost algorithm.For the existing method based on random hough transform,the method based on template matching is only applicable to the specific scene,the specific angle of the hammer detection,does not apply to the problem of this data set.In this paper,an anti-hammer detection method based on AdaBoost algorithm is proposed.During the training phase,we used the Haar feature for the damper and then trained with the AdaBoost classifier.In the detection phase,due to the relationship between the damper and the power line,we first extract the power line,select the area with the hammer of interest,and then through the sliding window with the classifier in this area for vibration hammer detection.Finally,we compare the influence of different feature selection and different parameter setting on the experimental results,and evaluate and analyze the experimental results from both accuracy and computational efficiency.In the field of UAV collected field images,we collected and collated two hundred and five hundred images containing damper,established a damper data set,and manually marked the earthquake hammer Ground Truth,and finally in this data set of detection The image was tested against damper,reaching 93%accuracy.(2)In view of the situation of false damper,we propose a method based on multi-angle matching to supplement the fault damper,so as to achieve the purpose of optimizing the test results.(3)Based on the deep learning damper detection.Aiming at the complicated scene of the images collected by UAVs in this subject,we propose a method to detect the damper based on the deep learning SSD model,and design the detection model.Finally,the experimental results are analyzed.On the data set we build,the SSD model achieves 98%accuracy.