基于 CMT 的红外可见融合跟踪算法

现有大部分融合跟踪算法都使用分类器型跟踪算法，这种算法在目标尺度发生变化时会存在一定误差。虽然已有一些解决方案，但仍不能从根本上解决该问题，而且一般无法估算目标的旋转状态。为此，基于CMT 算法提出一种新的基于点的多传感器融合跟踪算法。在获取目标状态上综合了多传感器信息，并对目标中心使用卡尔曼滤波进行修正，能够跟踪目标并获得目标的尺度及旋转信息。实验结果表明，该算法在常见分辨率下跟踪非旋转目标，重叠率与现有算法相当，在分辨率较高的情况下，在实时性上具有优势。

Infrared Visible Fusion Tracking Algorithm Based on CMT

Most of the fusion tracking algorithms use the classifier-type tracking algorithm. This algorithm has certain errors when the target scale changes. Although there are certain solutions，it can not be solved fundamentally，and generally can not estimate the rotation state of the target. In order to solve these problems，a new point-based multi-sensor fusion tracking algorithm based on CMT is pro-posed. The multi-sensor information is integrated on the acquisition target state，and the target center is corrected by using Kalman filtering，and the target can be tracked and the scale and rotation information of the target can be obtained. Experiments show that the algorithm tracks non-rotating targets at common resolutions and the overlap ratio is comparable to existing algorithms，and it has advantages in real-time performance at higher resolutions.