论文部分内容阅读
为保证无人机在视觉传感器的辅助下进行自主着陆的着陆精度、实时性和安全性,需要具备一套跟踪速度快、稳定性高的图像处理跟踪算法。本文首先介绍在检测识别到目标后,在目标所在区域网格撒点,初始化若干个跟踪特征点;之后,采用L-K光流法跟踪初始化后的像素点,并采用前后误差法,筛选去除跟踪效果差的点;然后,为保证跟踪目标的完整性,采用模板匹配,比较前后两帧跟踪的目标的相似性,以此结果判断是否需要重新检测识别;最后,以拍摄的视频资料为素材进行算法实时性和跟踪稳定性的验证。实验结果表明设计的目标跟踪算法可以快速稳定的跟踪目标,满足无人机进行自主着陆的要求。
In order to ensure the landing accuracy, real-time and safety of UAV assisted by visual sensor, it is necessary to have a set of image processing and tracking algorithms with high tracking speed and high stability. In this paper, firstly, after detecting and recognizing the target, we introduce a number of tracking feature points in the grid of the target area and initialize a number of tracking feature points. After that, the LK optical flow method is used to track the initialized pixel points and the tracking error Then, in order to ensure the integrity of the tracking target, the template matching is used to compare the similarity of the two tracking targets before and after, so as to judge whether the detection needs to be re-detected or not. Finally, taking the video data taken as the algorithm Real-time and tracking stability verification. The experimental results show that the designed target tracking algorithm can quickly and stably track the target and meet the requirement of autonomous landing of the UAV.