基于Apriltags改进算法的无人机移动目标识别与跟踪

摘要

移动目标识别与跟踪,在视频监控、人机交互、智能交通、军事应用等领域具有重大应用价值.本文针对当前目标识别与跟踪领域普遍存在的处理速度较慢、实时性不足等问题,提出了一种基于Apriltags识别的改进算法,对移动目标进行局部搜索,并结合Kalman滤波器实时估计目标下一时刻在图像中的位置,大幅提升了算法处理速度和跟踪性能.本算法在大疆M100四旋翼无人机平台上,搭载Manifold机载计算机完成了实验测试.实验证明,算法鲁棒性强、稳定性好,成功实现了无人机对快速移动目标的识别与稳定跟踪.%Moving target recognition and trackinghas a significant application value, in the field of video surveillance, human-computer interaction,intelligent transportationandmilitary applications. In order to solve with the problem of lowrecognition speedin the target recognition and tracking field , this paper presents an improved algorithm based on Apriltags recognition. It detects moving targetlocally with Kalman filter's real-time estimation of the target's 2D position in the image. It dramatically improves the algorithm processing speed and tracking performance. The algorithm is tested in the M100 UAV platform , which is equipped with Manifold. Experiments show that the algorithm has good robustness and good stability , and it successfully realizes the recognition and stable tracking of UAV for fast moving target.