多目标舰船自动跟踪方法研究

发布时间：2019-08-24 14:32

【摘要】：目前针对背景复杂的海上舰船作为跟踪对象的跟踪算法研究较少。另外,多目标跟踪算法在变化频繁的背景条件下的精确性和实时性不足。为了提高多目标舰船跟踪的视觉跟踪算法的实时性和鲁棒性,本文提出了一种改进的基于TLD(Tracking-Learning-Detection)多目标舰船自动跟踪算法。该算法实现了海上多个目标舰船准确实时跟踪。首先,算法应用海天线检测与提取方法,利用最大类间方差阈值分割和Hough变换提取海天线;其次,算法利用Kalman滤波原理对海天线上的目标舰船进行定位与检测,分离出目标舰船;最后通过提取出来的目标用其最小外接矩形生成目标初始跟踪框,跟踪器利用初始跟踪框的位置坐标信息对目标进行实时自动跟踪。算法运行过程中都在海天线附近进行扫描检测,缩小了图像遍历范围,提高了算法的实时性。实验表明,通过对比典型的Mean-shift算法以及原始TLD目标跟踪算法,本文算法跟踪结果的精度较高,实时性较好。
[Abstract]:At present, there is little research on tracking algorithm for marine ships with complex background. In addition, the accuracy and real-time performance of multi-target tracking algorithm under frequently changing background conditions are insufficient. In order to improve the real-time and robustness of the visual tracking algorithm for multi-target ship tracking, an improved multi-target ship automatic tracking algorithm based on TLD (Tracking-Learning-Detection) is proposed in this paper. The algorithm realizes the accurate and real-time tracking of multiple target ships at sea. Firstly, the sea antenna detection and extraction method is used, the maximum inter-class variance threshold segmentation and Hough transform are used to extract the sea antenna, secondly, the Kalman filtering principle is used to locate and detect the target ship on the sea antenna, and the target ship is separated. Finally, the extracted target uses its minimum external rectangle to generate the initial tracking frame of the target, and the tracker uses the position coordinate information of the initial tracking box to track the target automatically in real time. During the operation of the algorithm, scanning and detection are carried out near the sea antenna, which reduces the range of image traversing and improves the real-time performance of the algorithm. The experimental results show that by comparing the typical Mean-shift algorithm with the original TLD target tracking algorithm, the tracking results of this algorithm have higher accuracy and better real-time performance.
【作者单位】：华南理工大学自动化科学与工程学院;国家海洋局南海调查技术中心;
【基金】：海洋公益性行业科研专项经费(201505002) 中央高校基本科研业务费重点项目(2015ZZ028) 自主系统与网络控制教育部重点实验室资助~~
【分类号】：E925;TP391.41

【相似文献】