Research On Application Of Visual Tracking Technology Of Insulator For Transmission Line

With the rapid development of power transmission system,the traditional manual inspection methods are difficult to meet the needs of modernization construction and development,and the emerging unmanned aerial vehicle line-cruising method starts to be applied.In this paper,on account of the application scenarios of unmanned aerial vehicle electric transmission line inspection,it studies the insulator multi-scale tracking,to provide technical support for the intellectualization of unmanned aerial vehicle inspection in the future.The accurate estimation of changing in scale is a challenging problem in the tracking algorithm.Based on the spatio-temporal context tracking algorithm,it is difficult to adapt to the target of the insulator.In the paper,a tracking algorithm is proposed to realize the multi-scale tracking of insulators.The major research contents are:Firstly,the classical target tracking algorithm is generalized,and their principles and advantages and disadvantages are analyzed.Aiming at the tracking algorithm is difficult to adapt to the change in scale,the paper achieves the multi-scale tracking of insulators by means of selecting the combination of the spatio-temporal context and correlation filtering.Then,in the learning process of the spatio-temporal context model,only the grayscale features are used to describe the target,and the ability of the appearance of the target is weak,and the discriminant ability of the model is poor.The gray feature of single channel is improved to multi-channel HOG feature,and the HOG feature is reduced by principal component analysis(PCA)to improve the real-time performance of the algorithm.Secondly,in order to improve the spatio-temporal context algorithm can not adapt to the defect of insulator scale changes,inspired by the relevant filtering principle,the method of the spatio-temporal context and correlation filtering with the insulator tracking as target position estimation and target scale estimation of two independent parts.Firstly,the spatio-temporal context model is trained by the HOG feature,which is responsible for detecting the change of the target position.The probability value of each point of the sample is calculated by the spatio-temporal context model,and the target position is determined according to the maximum probability value;And then uses the HOG feature to train the scale correlation filter.After determining the target position of the insulator,the samples with different scales are extracted and matched with the scale filter.Finally,in the laboratory and field environment,it takes tempered glass insulator and ceramic insulator as the goal,to test the algorithm and analyze the overlap rate to verify the accuracy of the algorithm.The experimental results showed that the improved algorithm can track the insulator accurately and adapt to the change of scale.