基于直线和SIFT融合特征的物体识别研究  

Line and SIFT fusion feature based object recognition

作　　者：李永红[1] 卢昭金[1] 刘哲[1] 石海杰[1] 

机构地区：[1]中北大学信息与通信工程学院,太原030051

出　　处：《国外电子测量技术》2009年第6期31-34,共4页Foreign Electronic Measurement Technology

摘　　要：本文阐述了基于直线特征和SIFT(scale invariant feature transform)特征融合的物体识别研究,直线特征是最普遍的图像特征之一,而SIFT特征是最鲁棒的局部特征,其对旋转、尺度缩放、亮度变化保持不变性。本文以家用遥控器作为目标物体,首先提取平行直线特征作为本问中的主要图像特征,通过与物体模型进行匹配可以产生出很多的假设匹配结果,SIFT特征作为辅助特征,用来降低匹配的不确定性,从而提高识别率。试验结果表明该方法在复杂的环境中达到了鲁棒的识别效果。This paper explains the object recognition approach based on line and SIFT fusion feature, line feature is one of popular image features, and SIFT feature as the most robust local feature, which is invariant to rotation, scale, change of illumination. This paper uses remote control as the target object, parallel line pair （PLP）features are extracted firstly, which are the main feature for recognition, based on PLP, a number of hypothesis are generated, and then SIFT feature is applied as additional feature to verify those hypothesis in order to reduce the ambiguity. Experimental results show that our approach can achieve robust recognition result in the cluttered environment.

关 键 词：物体识别 特征融合 SIFT特征 直线特征 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术] TP391.41[自动化与计算机技术—计算机科学与技术]