Abstract: In order to improve the real-time performance of the visual navigation of autonomous aerial refueling with hose type, a adaptive kernel window algorithm based on Meanshift algorithm is proposed, which can improve the real-time performance of the algorithm. Taking the black circular area inside the drogue as the final tracking target, the detection algorithm provides the initial relative position of the drogue, accuratly and reliably; then it takes the position information of the drogue as the input of Meanshift algorithm to update the kernel window to realize drogue image target tracking. It has fast processing speed and high precision. The final results from the validity, rapidity and accuracy of detection and tracking methods are analyzed and compared. When the precision is not affected, the tracking processing time is reduced by about 30% than the detection time. Adaptive kernel window tracking algorithm proposed in this paper is fast and efficient, and it has the same accuracy as the global detection. Processing speed increases by 30% and reliability is above 95%.

Key words: global detection, adaptive kernel window, target tracking, circular target

摘要： 为提高软管式自主空中加油视觉导航的实时性，提出了一种基于Meanshift算法改进算法——自适应核窗口跟踪算法，用于锥套目标的跟踪，可大幅度提升算法的实时性。以加油锥套内部的黑色圆形区域为最终的跟踪目标，检测算法主要提供初始锥套的相对位置，检测准确可靠；然后将锥套的位置信息作为自适应核窗口跟踪算法的输入以更新跟踪目标，实现锥套图像目标的跟踪，处理速度快，精度满足要求。最终实验结果分别从有效性、快速性以及准确性对检测和跟踪方法进行了分析对比，在基本上不影响精度的情况下，跟踪处理时间比检测时间降低了30%左右，效果显著。提出的自适应核窗口跟踪算法快速有效，精度与全局检测精度相差无几，处理速度提升了30%，可靠性在95%以上。

关键词: 全局检测, 自适应核窗口, 目标跟踪, 圆形目标