作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2012, Vol. 38 ›› Issue (19): 191-194. doi: 10.3969/j.issn.1000-3428.2012.19.049

• 人工智能及识别技术 • 上一篇    下一篇

基于区域CSIFT特征的目标识别算法

罗 磊a,范彩霞b   

  1. (西安理工大学 a. 后勤管理处;b. 印刷包装工程学院,西安 710048)
  • 收稿日期:2011-12-08 出版日期:2012-10-05 发布日期:2012-09-29
  • 作者简介:罗 磊(1982-),男,工程师,主研方向:图像处理;范彩霞,讲师、博士研究生

Object Recognition Algorithm Based on Region CSIFT Feature

LUO Lei   a, FAN Cai-xia   b   

  1. (a. Department of Logistics Management; b. Faculty of Printing and Packaging Engineering, Xi’an University of Technology, Xi’an 710048, China)
  • Received:2011-12-08 Online:2012-10-05 Published:2012-09-29

摘要: 无重叠多摄像机监控系统在使用局部形状特征进行目标识别时,会忽略颜色信息,且对光学变换具有不稳定性。为此,提出一种基于区域彩色尺度不变特征变换的目标识别算法。通过修正的双色反射模型,提取对光照特性和物体几何特性具有不变性的颜色特征,得到尺度不变特征变换描述子,利用颜色特征和形状特征建立目标模型。实验结果表明,该算法对刚体和非刚体目标的识别都能取得较好的效果。

关键词: 多摄像机, 目标区域检测, 目标识别, 反射模型, 颜色不变量

Abstract: In non-overlapping multi-camera surveillance system, using the local shape features for target identification is unstable to optical transformation and ignores the color information. To solve this problem, an object recognition algorithm based on region Colored Scale Invariant Feature Transform(CSIFT) is proposed. The color feature is extracted by modifying dichromatic reflection model, which is invariant to light conditions and geometry character, the Scale Invariant Feature Transform(SIFT) descriptor is generated on the color invariant feature. The object appearance model is established by color and shape feature. Experimental results show that this algorithm can achieve a better match results for rigid and non-rigid object recognition.

Key words: multi-camera, object region detection, object recognition, reflection model, color invariant

中图分类号: