基于TLD算法的加油锥套跟踪

发布时间：2017-12-28 08:32

本文关键词：基于TLD算法的加油锥套跟踪　出处：《应用光学》2016年03期 　论文类型：期刊论文

【摘要】：针对自主空中加油对接阶段锥套跟踪问题,提出了一种基于tracking-learning-detection(TLD)的锥套跟踪算法。该算法将加油锥套的跟踪任务分解成跟踪、学习、检测3个部分。跟踪模块在LK光流法的基础上添加跟踪失败自检测,筛选出好的跟踪点,跟踪加油锥套;检测模块构建级联分类器,对滑动窗遍历得到的图像块进行分类并返回含有目标的图像块,融合跟踪模块的跟踪框,给出最终跟踪结果;学习模块引入P-N约束修正错误样本并学习更新检测模块。利用Creator/Vega Prime软件对空中加油进行视景仿真,在视景仿真视频上测试锥套跟踪算法。结果表明:TLD算法跟踪加油锥套成功率达95.5%,处理每帧平均耗时31.4ms,能够满足加油锥套跟踪鲁棒性、准确率、实时性的要求。
[Abstract]:A conical sleeve tracking algorithm based on tracking-learning-detection (TLD) is proposed to track the conical sleeve tracking in the autonomous air refueling docking stage. The algorithm decomposes the tracking task of the refueling cone into 3 parts: tracking, learning and detecting. The tracking module add tracking failure self detection based on LK optical flow method, selection of good tracking, tracking the refueling drogue; construction of cascade classifier to classify the image detection module, block sliding window traversal obtained and returns the image block containing target tracking module, fusion tracking frame, gives the final tracking result; learning the introduction of P-N module constraint error samples and study the change detection module. The Creator/Vega Prime software is used to simulate the aerial refueling, and the cone tracking algorithm is tested on the visual simulation video. The results show that the success rate of TLD algorithm tracking fuel cone is 95.5%, and the average time consuming for each frame is 31.4ms, which can meet the requirements of tracking robustness, accuracy and real-time.
【作者单位】：空军工程大学航空航天工程学院;
【基金】：国家自然科学基金资助项目(61473307) 空军工程大学优秀博士论文扶持基金(KGD081114006)
【分类号】：TP391.41;V279
【正文快照】： 引言无人机的出现深刻地改变了现代作战方式,但受携带燃油量限制,其作战效能得不到充分发挥。而自主空中加油技术可以延长无人机留空时间,扩大作战半径,增加有效载荷,充分发挥了无人机的优势[1]。目前空中加油主要有硬式(探针式)和软式(插头-锥套式)两种方式。硬式加油速度快

【相似文献】