特征一致红外弱小目标匹配与定位研究

Research on matching and locating of infrared dim small targets with the same characteristics

应用基于双目立体视觉原理的自动目标定位系统,研究特征一致的红外弱小目标且目标数量不唯一的情况下的,目标匹配与定位。提出了基于时间序列分割-极线约束的匹配算法。时间序列分割,形成若干帧段,对每个帧段目标单独处理,缩小匹配范围。同个帧段内,利用极线约束条件,即从目标空间分布角度匹配。初步解决系统无法在毫秒量级同步,以及同一帧图像出现多个目标造成的难点。对现场实际采集图像序列的15个目标进行时间序列分割匹配,对外场模拟实验同帧5个目标进行基于极线约束匹配。实验结果证明,算法提高了多个红外弱小目标匹配的准确率。

With the automatic target locating system based on binocular stereo vision principle, matching and locating of infrared dim small targets with the same characteristics and the number of targets is not only one is researched. The matching algorithm based on time series segmentation and polar constraint is proposed. Time series segmentation formats a number of segments, and the matching is for each segment, thus the scope of matching is narrowed. In the same segment, the matching is based on polar constraint from the perspective of spatial distribution. The difficulty that the system cannot be synchronization in milliseconds order is preliminary solved, and the multiple targets are in the same frame. 15 targets in live images are matched with time series segmentation, and 5 targets in the same external field simulation image are matched with polar constraint. The experimental results demonstrate that the algorithm improves the accuracy in multiple infrared dim small targets matching.