双色红外图像目标多特征决策级融合识别算法

Dual band IR image target multi-features decision level fusion recognition algorithm

针对双色红外成像系统中的自动目标识别问题,提出了一种采用多特征多分类器决策级融合的目标识别算法。该算法首先提取目标的形状特征和面貌特征;接着基于各种不同特征设计多个分类器对目标进行分类;然后采用所设计的多分类器决策级融合策略对多个分类器的目标分类结果进行融合处理;最后采用所提出的决策规则对多分类器融合分类结果进行处理得到最终的目标识别结果。该算法充分利用了目标在多传感器图像中的多种分类特征信息,在较大程度上提高了系统的目标识别效率和精确性。实验结果证实了该算法的有效性。

Aiming at the problem of ATR for the dual band IR imaging system,a method of target recognitionis proposed using multi-features and multi-classifiers decision level fusion.The algorithm firstly extracts shape features and appearance-based features;Following,multiple classifiers are devised based on the different features to classify the target;Then,the target classification results of multiple classifiers are fused using proposed strategies of multi-classifiers decision level fusion;Finally,the multi-classifiers fusion classification results are deal with using devised decision rules to get the system＇s target recognition results.The method makes good use of the multiple kinds of classification feature information of target in multi-sensors images,improves the performance and accuracy of target recognition to some degree.Experimentation results demonstrate the effectiveness of the proposed algorithm.