基于多特征融合的红外目标关联算法  

Infrared target association base on multi-feature fusion

作　　者：张进[1] 魏敏[1] 卢宇[1] 吴钦章[1] 

机构地区：[1]中国科学院光电技术研究所

出　　处：《红外与激光工程》2008年第3期551-555,共5页Infrared and Laser Engineering

基　　金：中国科学院创新工程资助项目(H06k075)

摘　　要：利用红外目标同时具有位置、灰度、面积等多特征的特点,提出了一种基于多特征融合的目标关联算法。首先在极坐标系下对目标位置采用概率数据关联算法计算候选目标的关联概率,然后结合目标的灰度、面积特征的预测误差计算关联波门中的候选目标在各种特征条件下的关联概率,进而利用多特征融合方式,计算出综合关联概率,完成目标状态估计的更新。实验仿真结果表明,由于跟踪关联概率由多种特征共同确定,避免了目标位置特征信息不稳定所造成的跟踪精度下降的问题,实现了密集杂波环境下红外目标稳定跟踪,其跟踪精度和稳定性明显高于依靠位置特征信息进行关联的传统概率数据关联算法。An algorithm based on information fusion of multi-feature in infrared target tracking was proposed. It made use of the multiple-feature of the infrared target such as kinematics state, gray, and size. The association probabilities for targets position were calculated by using the probabilistic data association （PDA） in the polar coordinate system. Then the prediction errors of gray and size calculated separately according to state predictions and measurements were used to compute related association probabilities of targets. Finally, the decision of synthetic data association of all the targets in the validation region was made according to the information fusion of association probabilities with multiple-feature. Experimental results show that the proposed algorithm solves the problem of tracking precision decreasing while kinematics state of target is unstable. It is proven to be effective in infrared target tracking in heavy clutter. Compared with the traditional PDA algorithm, the new algorithm can obtain higher accurate and improve the stability of the infrared target tracking system.

关 键 词：目标跟踪 红外探测器 多特征融合 概率数据关联 

分 类 号：TN216[电子电信—物理电子学] TP391[自动化与计算机技术—计算机应用技术]