多步长帧间匹配的弱点目标跟踪算法

Algorithm of Multi-step Adjacent Frame Matching Used for Point Targets Tracking

针对海空复杂背景下的红外点目标检测跟踪问题,提出了一种基于多步长帧间匹配的点目标轨迹检测与跟踪算法。算法独特之处在于无需提前假定疑似目标点位置,采取对相邻两帧间像素点匹配,单个匹配过程与当前相邻帧外的序列帧无关,实时标记和更新匹配结果兴趣区域像素值;顺序输入序列帧,执行帧间匹配操作,同步形成目标运动轨迹。算法结构简单、存储量少、计算速度快,匹配过程不随运动目标点多少或运动状态变化而改变,即对不规则运动的多目标、交叉目标、迎面目标的匹配过程和对单一目标匹配过程完全相同,适应性强。根据多次仿真和实拍照片实验,证实了理论上区别于传统算法的上述优点,有较高的应用价值。

Aiming at the detection and track of IR point targets in complicated sea-sky background, this article presents an algorithm of targets detecting and tracking based on multi-step adjacent frame matching. The algorithm unique merit lies in no need to suppose, the doubtful target point in advance; the current match process has nothing to do with other frames. In the match result, we can obtain the interest region pixel point, input the pretreatment binary images gradually, perform matching operation, and form real time target trajectory synchronously. Especially, the match process is not changed by the moving state or the number of targets, and the match process has no different between a single target and multi-targets, cross-targets, headon targets. According to the simulation and the real images experiment time after time, we have already confirmed the abovementioned merits different from the traditional algorithms.