Development Of UAV's Visual Tracking Obstacle Avoidance Method Based On Embedded Platform

In recent years, the vision navigation technology of unmanned aerial vehicle(UAV) has caught widespread concern. At the same time, the microprocessor based on the embedded system is also showing its superiorities in the field of control engineering for self-reliance and cheap-price. In this paper, an embedded system is developed to solve the problem of UAV's obstacle avoidance in an unknown environment, mainly discusses the navigation obstacle avoidance strategy for UVA's flight,obstacle recognition and ranging. And then complete the design of algorithm and build hardware platform, in the end of paper, one experiment is done to verify the engineering value of the system.First, to find a proper navigation obstacle avoidance strategy, this paper improves the potential field function based on the traditional artificial potential field method. And then design a controller by combining the theory with the optimal control to guide the flight of the UAV. Under the control of the system, the UAV can avoid the obstacle autonomously and finally arrive at the target location.Then, for obstacle detection, the image pre-processing algorithm is studied, and on the basis of several classical optical flow algorithms, an improved pyramid L-K optical flow method is developed based on the corner detection, which uses the corners as feature points to compute the optical flow vector of the target scene. For ranging, use the optical method to find the obstacle's warning area, and then through the camera calibration, establish the corresponding mode of the two dimensional pixel to3 D world to compute the exact distance between UAV and obstacle.Finally, to build the embedded platform, complete the overall design of the system and the hardware implementation of the algorithm. The system designed in this paper includes two parts: the first is image acquisition module, which is based on CMOS image sensor, for collecting the image data in front of the UAV. The second is image processing module by using the FPGA+DSP architecture, in this part, the FPGA is used for image pre-processing and the DSP is used for image processing and outputting control signal. The results of the experiment show that the system has a good ability to detect the obstacle and plan the flight path, which means the algorithms developed in this paper are feasible and effective.