Research Of Detection And Recognition Of Obstacles On Transmission Lines

Ice covered on high voltage transmission lines often leads to abnormal and unstable conditions for power transmission lines to work properly. Incidents of power supply system paralysis caused by thick ice covered on high voltage transmission lines occurred in many nations. They also suffered huge economic loss. Hence, how to effectively and safely eliminate ice on power transmission lines has become the pressing need for the power system in many nations. With the constant development of robot technology, replacing traditional deicing methods with deicing robots has become an important research subject for solving the problem of ice on power transmission lines.Deicing robots work on high voltage transmission lines covered with ice. They have to walk, cross obstacle and eliminate ice on power transmission lines. While walking on power lines, how to accurately identify obstacles on power transmission lines is the necessary precondition for deicing robots to work properly. It is also a core project in the research area of deicing robots technology. Hence, according to the characteristics of obstacles on 220KV transmission line, a series of methods for obstacle vision detection and recognition are put forward in this thesis.The dangers caused by thick ice on high voltage transmission lines to people's life and work are introduced in the paper. Based on the summary concerning traditional ways of deicing, the significance of using deicing robots is also proposed as a new way of eliminating ice. The research status of power lines inspection robot and deicing robots in China and other countries abroad are summarized in the issue, as well as research achievement and development orientation for deicing robots to detect and identify obstacles. Based on the analysis of structural features of power lines, the mechanical structures and visual systems of three-arm deicing robots, and different types of obstacles on power lines are mainly discussed in this paper.Concerning the detection and identification of obstacles on power lines, following research is conducted in the paper:1. Considering the complexity of detecting obstacles on power lines, based on edge figures of images, the detection algorithm based on bounding box is proposed in the paper. By calculating the number of edges for images totally included the bounding box, obstacles are detected. When image preprocessing is conducted on the obstacles detected, noise jamming is reduced and readability is enhanced in the images. Based on structured random forests edge detection algorithm, select the optimal threshold value and extract images'edges. By defining the scoring functions of bounding box, conduct scoring on bounding box and detect them included in obstacles.2. As for the complexity of obstacles recognition, obstacles recognition algorithm based on convolutional neural network is proposed. First, the paper introduces the development background and history of convolutional neural network. Then, based on the basic structure of convolutional neural network, network architecture suitable for obstacle detection is proposed. Lastly, the network is trained based on the combination of forward scatter propagation and counter propagation, so as to achieve the classification and identification of obstacles.