基于混合特征提取的易混淆三维目标识别研究  

Researches On Seemingly Similar 3-D Target Recognition Based On Mixed Features Extraction

作　　者：王松伟[1] 薛淑磊[2] 

机构地区：[1]西北工业大学航天学院,陕西西安710072 [2]西安工业大学北方信息工程学院,陕西西安710032

出　　处：《计算机测量与控制》2008年第7期1036-1039,共4页Computer Measurement &Control

基　　金：国家自然科学基金资助项目(60575013);航空科学基金资助项目(04I53067);航天科技创新基金资助项目(05C53005)

摘　　要：针对易混淆三维目标识别这一问题,通过提取混合特征完成识别;首先获取各类三维目标的若干二维视图,将这些视图放在一起提取它们的大体轮廓特征并进行分类,使得具有相似外形的目标分为一类;然后采用基于统计学的标志矩提取方法提取每类中能分辨各种目标的局部特征,这样特征库就建立了;识别过程中,针对实际要识别的目标,提取它的混合特征并与库中数据进行匹配,确定其类型;算例研究表明,基于混合特征提取的识别系统结构简单、高效。Aiming at seemingly similar 3--D target recognition, a method of mixed features extraction is presented. Firstly, many 2--D viewers of every kind of target are acquired; putting all these views together, we extracts their Contour features under coarser resolution and clusters them into several classes; using flag--moment extraction method based on statistic theory, system extracts the detail features of every class under finer resolution which can distinguish different kinds of targets in this class. In this way, we have established the 3--D targets feature--lib. When,real--time target arrives, its mixed features are extracted and matched with data in feature--lib, and thenthe recognition result is obtained. The experiments show the simplification, efficiency of this system.

关 键 词：易混淆三维目标识别 混合特征提取 局部特征 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]