Multi-scale Image Tracking Technology Research Of Insulators

With the high-speed development of the national power grid, power inspection tasks also increase heavily, traditional way of power inspection with human is inefficient and insecure. In recent years, a new way of electric power inspection-unmanned aerial vehicles has been used and widely studied. This new technology provides powerful technical guarantee for the obstacle avoidance, distance measurement and intelligent inspection of power inspection.Based on the analysis of the requirement of unmanned aerial vehicle(uav) in power inspection, it takes the insulator as the research object, it puts forward on a kind of dimension self-adaptation insulator tracking algorithm. In this paper, the main work includes the following aspects:(1) With the analysis of international research status and technical difficulties of unmanned aerial vehicle in power inspection, according to the characteristics of uav in power inspection, considering the effect and accuracy of the real project, it selects a tracking algorithm based on compressed sensing as the basic algorithm.(2) As for the glass insulator tracking problems, it proposes a multi-scale insulator tracking algorithm. It uses a rescaling strategy based on compressive tracking to solve the multi-scale tracking problem.Adopting way of sparse projection to make dimensionality reduction on original image, it uses Lab space characteristics of insulator to make segmentation and changes scale of tracking box according to effective pixel percentage of insulator to solve the fixed search box problem in Compressive Tracking.(3) As for the ordinary insulator tracking problems, using particle filter method combined with Compressive Tracking to solve the multi-scale tracking problem. The method does not depend on the insulator segmentation. A 2-order transition model that considers the two previous statuses have been used to estimate the current position and scales tatus for each particle. Finally, using the na?ve Bayesian classifier response to ensure the target location and scale.(4) The test result indicates that this algorithm can make change in the scale variations of insulator self-adaptively in the lab and field environment.It uses qualitative analysis and quantitative analysis to analyze the deficiency of proposed algorithm which can provide guidance on the research in the future.