摘要
针对无人机视角下拍摄物体尺寸小、外观模糊,以及无人机平台自身存在运动等问题,基于无人机航拍视频的运动目标检测与跟踪仍具有很大的挑战性。提出了基于改进LK光流法的运动目标检测与跟踪算法,利用卡尔曼滤波算法进行目标跟踪;针对传统LK光流法在动态环境下效果不佳的问题,引入语义分割线程筛选动态点和静态点,通过筛选后的静态点求解单应矩阵完成无人机的运动补偿、计算动态点的光流值获得光流点以降低误匹配率,最后通过聚类算法和形态学运算得到完整的运动目标;将所提的重识别模块引入卡尔曼滤波算法实现目标跟踪。实验结果验证了所设计的运动目标检测与跟踪算法在无人机视角下能够实时、准确地提取出运动目标,并持续稳定地跟踪目标。
In view of the problems of small size and blurry appearance of images shot from the perspective of UAV and motion of the UAV platform itself the detecting and tracking of moving objects based on UAV images is still very challenging.A moving object detecting and tracking algorithm based on the improved LK optical flow method is proposed and the Kalman filter algorithm is used for object tracking.In order to solve the problem that the traditional LK optical flow method is not effective in the dynamic environment the semantic segmentation thread is introduced to screen the dynamic points and static points the motion compensation of the UAV is completed by solving the homography matrix through the static points and the optical flow value of the dynamic points are calculated to obtain the optical flow points for reducing the mismatch rate.Finally the complete moving object is obtained through the clustering algorithm and morphological algorithm and the proposed re-identification module is introduced into the Kalman filter algorithm to realize object tracking.The experimental results verify that the moving object detecting and tracking algorithm designed here can accurately extract the moving object in real time from the perspective of the UAV and track it continuously and stably.
作者
钱奕舟
王在俊
高耀文
王雪
QIAN Yizhou;WANG Zaijun;GAO Yaowen;WANG Xue(CAAC Academy of Flight Technology and Safety Civil Aviation Flight University of China,Guanhan 618000 China)
出处
《电光与控制》
CSCD
北大核心
2024年第11期47-54,共8页
Electronics Optics & Control
基金
民航飞行技术与飞行安全重点实验室自主研究项目(FZ2022KF02)
民航飞行技术与飞行安全重点实验室开放基金项目(FZ2020KF07)
2023年中央高校教育教学改革专项,科研基地创新创业基地建设(E2023065)
2023年中国民用航空飞行学院研究生创新基金项目(24CA FUC10175)。
关键词
无人机
LK光流法
运动补偿
运动目标检测
目标跟踪
UAV
LK optical flow method
motion compensation
moving object detection
object tracking
