中远距离舰船目标识别方法研究  

Method for ship target recognition in mid-long range

作　　者：赵峰民[1] 刘皞[1] 陈望达[1] 

机构地区：[1]海军航空工程学院科研部,山东烟台264001

出　　处：《激光与红外》2012年第9期1071-1075,共5页Laser & Infrared

基　　金：装备预先研究项目(No.51301010102)资助

摘　　要：分析了中远距离舰船目标的特点和对其进行识别的重要性,提出了一种新的舰船目标识别算法。该算法采用基于曲线进化的方法提取目标的骨架,通过计算目标骨架扩展的Hu不变矩提取目标特征,最后选用支持向量机作为分类器,并用创建的舰船目标的模型对分类器进行训练,利用训练好的分类器对实际获取的目标的红外图像进行分类识别。实验结果表明算法的识别率较高,鲁棒性较好。Characteristics and importance to identify ship kind target in mid-long range are analyzed, and a new ship target recognition algorithm is proposed. Firstly, the skeleton of target is extracted by a method based on curve evolution. Then the Hu invariant moment of the target skeleton is calculated as the feature of target. Finally, SVM is selected as the classifier. The classifier is trained by the ship model created and the real target is recognized by the well trained classifier. The exper/menta/result shows that the method is effective and robust.

关 键 词：目标识别 骨架图 HU不变矩 支持向量机 

分 类 号：TN216[电子电信—物理电子学]