一种基于多传感器多级信息融合的红外目标检测方法

针对远距离低信噪比条件下目标检测难的实际问题，该文提出一种基于多传感器多级信息融合的目标检测方法。该算法包括两个部分：特征级融合和决策级融合。在特征级融合阶段，首先提取双波段红外图像的各个特征图像；然后，采用自适应加权方法对所得到的特征图像进行融合，得到目标判决置信图；最后根据置信度最大规则对置信图进行目标位置搜索，得到各级目标判决结果。在决策级融合阶段，对各级目标判决结果采用组合逻辑进行融合，得到系统的目标检测输出。实验结果显示了该算法的有效性。

IR Target Detection Based on Multi-sensor Multi-level Information Fusion

Aimed at the difficult problem for detecting distant small target with very low SNR, the proposed method of target detection based on multi-sensor multi-leveI information fusion consists of two parts: The feature level fusion and the decision fusion. On the phase of feature level fusion, first extract all feature images of dual band IR images; then fuse these feature images by using the adaptive weighting method to get the confidence images of target decision; finally scan the confidence images by using the maximum confidence value rule to get result of target detection for all levels. On the phase of decision fusion, fuse the result of target detection for all levels by using the combination logic to get target detection output for the system. The result shows the effectiveness.