基于兴趣点多特征融合的物体识别方法  

Object Recognition Method Based on Fusing Multi-features of Interest Points

作　　者：李伟生[1] 赵灵芝[1] 

机构地区：[1]重庆邮电大学计算机科学与技术研究所,重庆400065

出　　处：《计算机工程》2010年第18期7-9,共3页Computer Engineering

基　　金：国家自然科学基金资助项目(60842003)

摘　　要：提出一种基于兴趣点多种特征融合的物体识别方法。利用简化的局部二值模式算子去除Harris冗余角点,提取感兴趣区域的3种特征并加权融合特征,在K最近邻(KNN)方法中引进加权因子计算特征距离函数,得到合适的分类器。实验结果表明,该方法能有效提高物体识别的正确率。This paper proposes a method for object recognition based on fusing multi-features of interest points. It uses Harris to detect corners, and then uses a simplified Local Binary Patterns（LBP） to wipe out some redundant corners. It gives a weight to every feature according to that it contributes to every class of object. In the K-Nearest Neighbors（KNN）, it introduces weight of feature to distance function to achieve a classifier adapted to every object. Experimental results show that this method effectively improves the object recognition accuracy.

关 键 词：物体识别 兴趣点 局部二值模式 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]