基于视觉的无人机感知与规避系统设计

结合无人机自身携带的光电平台,设计了一个面向空中动态目标的无人机感知与规避系统。使用图像阈值法结合帧差法检测出运动目标,利用Mean shift与Kalman滤波融合,在图像中实时跟踪目标;利用自抗扰控制器控制无人机光电平台光轴指向目标并进行激光测距以实现目标定位;最后采用人工势场法,实时地计算下一时刻航路点,对目标进行规避。仿真实验表明：针对虚拟现实产生的模拟视频流,系统所采用方法具有良好的实时性和有效性。

Vision-based System Design for UAV Target Detection and Avoidance

With the photoelectric platform on UAVs, a system for aerial moving target detection and avoidance is designed. First, image thresholding is used to reduce the impact of the complicated background. Then, the moving target is detected through frame difference. To tracking fast and robustly from UAVs, Mean shift search fusing Kalman prediction is used. After that, auto disturbance rejection controller is applied to control the photoelectric platform keeping the axis tracking the target. When the target remains in the center of the picture, laser distance measuring is used for target location. Finally, the artificial potential field method is applied to calculate the next path point and avoid the target. The experimental results present effectiveness of the approaches applied in the system.