Research On Key Technology Of Millimeter - Wave Radar For Power Line Detection

To ensure the safety of helicopters, a lot of researches have been conducted in millimeter-wave radar collision avoidance system. An important and difficult part in the radar system is to detect the high-voltage power lines as power line itself is very difficult to be detected and lots of helicopter crash accidents were caused by power lines. In this dissertation, we are aiming to detect high-voltage power lines using MMW radar, and the main works in this paper have been done as follows:1. We elaborate the principle of ranging and basic signal processing in MMW radar system based on Unmanned Aerial Vehicle (UAV) Obstacle Avoidance System (OAS) which has been complete.2. We establish a detail model of power line and simulate the RCS of power line in the millimeter-wave band. Then we obtain the power lines’ Bragg pattern and analyze the electricity scattering mechanism of power line.3. A method of mechanical scanning millimeter wave radar imaging system has been studied. Based on this system, simulations of imaging using the Bragg pattern have been done. The power lines appear as parallel lines in the radar video after a coordinate transformation, and Hough transform can be employed to detect them. The major challenge is that the videos are exceptionally noisy due to the ground return, and noise points could fall on the same lines which results in signal peaks after the Hough transform which are similar to the actual power lines.4. To differentiate the power lines from the noise lines, we propose some machine learning methods to classify the candidate lines which have been detected after Hough transform. In this dissertation we resort to three supervised machine learning methods, which are Decision Tree, Naive Bayes and Support Vector Machine. As supervised machine learning methods, lots of training data sets will be needed. In this paper, we analyze the features of power lines and noise lines in the radar video and propose ten dimensions vector as the feature vector which are used to train the classification. Simulation results demonstrate the effectiveness of the proposed algorithm. Finally, a detailed comparison of the three algorithms is done.