基于多特征融合的弱小运动目标识别

利用单个特征识别强噪声中的弱小运动目标,常因所提取的目标特征与噪声特征易混淆而导致高的虚警率.提出一种新的基于多特征融合的弱小运动目标识别方法.分析了弱小运动目标的连续相关性、面积及质心位置偏移这三个特征的可靠性及提取方法,对获取的特征值进行归一化后采用多特征融合的方法构造更具有鲁棒性的联合特征,确定了以具有最大多特征融合值为真实目标的决策方法.通过与采用单一特征的目标识别方法进行比较,证明了提出的多特征融合方法能更准确地识别弱小运动目标.

Small moving target recognition based on multi-feature fusion

Recognition algorithms for small moving target in strong noise based on single feature has a high false alarm. Sometimes the features of target and noise are very alike. A new recognition algorithm of small moving target based on multi-feature fusion is presented. The reliability and the extraction method of the continuous correlation, area and centroid position of the small moving target are analyzed. The normalized features are used to construct robust combined feature that can present the real target based on the multi-feature fusion method. The candidate target with the largest value of fusion feature as the real target is determined. The comparison with recognition method based on only one feature proved that the recognition based on multi-feature fusion could exactly recognize the small moving target.