相邻帧间匹配的迎头点目标跟踪算法

针对海空复杂背景下迎头点目标检测与跟踪难题,提出了一种基于相邻帧间匹配的边检测边跟踪算法.算法对相邻红外图像序列帧间点与点的邻域匹配,标记匹配结果兴趣区域像素点,统计标记次数,与输入单帧图像同步显示迎头目标检测结果.算法主要特点在于无需提前假定疑似目标点位置,单个匹配过程与当前相邻两帧外的其它序列帧无关,整个匹配过程不随目标数目多少或运动状态变化而改变.根据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,在军事应用中具有较高的参考价值.

An Algorithm of Adjacent Frame Matching Used for Head-on Point Targets Tracking

Aiming at the detecting and tracking of head-on point targets in complicated sea-sky background, an algorithm that could detect while tracking is proposed based on adjacent frame matching. The algorithm obtains the interest region by the adjacent frame matching in infrared image sequences, adopts a mark image to memorize the match result, adopts a statistical image to count the marking times of each pixels and outputs the statistical image to show head-on targets with input images. The algorithm is unique in no need to suppose the doubtful target points in advance; the single match process has nothing to do with other frames, and the whole match process does not change with the target movement state or the numbers of targets. Especially, the algorithm solves the head-on targets detection problems under the highly noisy condition. According to the simulations and experiments under the different condition time after time, it has been confirmed that the advantage above with high application value is different from the traditional algorithms and this method is of feasibility and validity.