Vision-Based Flight Control System for Unmanned Air Vehicles

• Main Authors: Lin zhen-yan, 林貞言
• Other Authors: 瞿忠正
• Format: Others
• Language: zh-TW
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/83259059208101568961

碩士 === 國防大學理工學院 === 電子工程碩士班 === 97 === In recent years, autonomous flight and navigation control, key technologies in Mini-UAVs or MAVs, have been widely utilized in military and civilian applications. Though MEMS gyros are widely used in flight control, they are not accurate for Mini-UAVs or MAVs. Therefore, infrared or vision sensors have become an alternative solution in flight control. Since the vision seneor is the basic equipment for most UAVs and MAVs, using the video camera is a good solution based on low cost and payloads. In this paper, we develop a vision-based flight control and obstacle avoidance system. Using the horizon line detected from image frames, we can calculate two important flying parameters, the roll angle( ) and the pitch value( ), which are dependant on the gradient and the position of the horizon line respectively. First the image is segmented into sky and ground regions by the horizon line detection algorithm, and the obstacles are detected in sky region by the obstacle detection algorithm. Then, we can obtain the flying parameters to avoid the obstacle and keep flight. We develop a real-time, robust, and accurate horizon-detection algorithm, whose detection rate is over 98%, and present an obstacle-detection algorithm. We use a remotely-piloted aerial vehicle which is equipped with a small CMOS video camera to transmit the image to the ground-based computer. Then, the computer uses the received images to detect the horizon and to estimate the roll angle and pitch value. In order to get the voltages sent from remote controller to modify the flying attitude, we take the PI control mechanism to transfer the flying parameters to the corresponding DC voltages. When a obstacle is detected, the vehicle can avoid the obstacle automatically. We not only propose a robust horizon and obstacle detection algorithm, but also complete a practicable system. Combining with image capture card, D/A converter card, and remote controller of UAV, we have implemented many vision-based flight control and obstacle avoidance tests. The experimental results are demonstrated with different environments such as sunny day, cloudy day, and interfering image.