旋翼无人机实时视觉跟踪算法

Real Time Visual Tracking Algorithm for Rotor Unmanned Aerial Vehicle

针对背景变化、目标状态变化和短暂遮挡导致的无人机跟踪失败问题,提出了融合基于加速鲁棒特征的光流跟踪器、基于归一化相关匹配(wormalized cross correlation,NCC)的检测器和基于卡尔曼滤波器的位置预估器的目标跟踪算法.此外,设计了基于串级比例-积分-微分控制器(proportion integration differentiation,PID)的位置-姿态外内环控制器,计算目标相对无人机的水平位移,据此通过两级PID算法完成位置-姿态的控制,以确保目标始终位于图像中心附近,实现对目标的有效跟踪.测试结果验证了该跟踪算法的有效性.

In response to the change of background, the change of target state and transient occlusion could result in tracking failure, an automatic target tracking algorithm of the rotorcraft unmanned aerial vehicle（UAV） was designed and implemented. It involved the target tracking in image sequences and UVA flight control, it also integrated optical flow tracker based on speeded up robust features （SURF）, detector based on Normalized Cross Correlation matching and position predictor based on Kalman filter. Besides, a position-attitude outer-inner loop controller based on cas- cade proportion integration differentiation （PID） was designed. The horizontal displacement of target relative to the UAV was computed as control input, and the position-attitude was adjusted through a two-step PID algorithm to en- sure the target is near the image center in order to achieve an effective tracking. The results confirmedits effectiveness.