搜索区域和目标尺度自适应的无人艇海面目标跟踪

海面目标跟踪任务是实现水面无人艇自主化航行、智能化作业的重要基础。相比于普通场景的目标跟踪,海面目标跟踪需要面对目标抖动剧烈及目标尺度变化大等问题。针对海面目标在图像画面中抖动剧烈的问题,本文提出了搜索区域自适应算法,该方法通过对海面场景的分割完成了海天线位置的提取,然后通过海天线运动模型自适应地确定了每帧图像中目标搜索的区域;针对跟踪过程中海面目标尺度变化较大的问题,本文通过分割搜索区域的方法实现了目标尺度变化的自适应跟踪。基于相关滤波跟踪框架并结合上述两种改进策略,在真实的海面目标图像测试序列中,本文算法相比传统的相关滤波算法在跟踪精度上至少提升了26%,有效地解决了目标抖动剧烈和尺度自适应问题,提高了海面目标跟踪任务的精度。

Search area and target scale adaptive sea surface object tracking for unmanned surface vessel

Sea target tracking is important in the autonomous navigation and intelligent operation of unmanned surface vessels. Compared to target tracking in common scenes, sea surface target tracking faces unique challenges, such as intense dithering of the target and considerable changes in its scale. Aiming at the problem of intense dithering of sea surface targets, an adaptive search area algorithm was present in this paper. The proposed method extracts the position of the sea-sky-line by segmenting sea surface scenes, and it adaptively determined the target’s search area in each frame using the motion model of sea-sky-line. To solve the problem of considerable changes in the scale of sea surface targets, this study achieved an adaptive tracking of the targets’ scale by segmenting the search area. Based on the correlation filtering tracking framework, combined with the two improves strategies above, the proposed algorithm improved tracking precision by at least 26% compared to the traditional correlation filtering algorithm. Therefore, the proposed algorithm effectively solves the problem of intense object jitter and scale adaptation and improves the accuracy of sea object tracking.