一种不同光照条件下的变动物体检测方法  

A Method for Change Object Detection under Different Illumination

作　　者：李亚兰[1,2] 邱永生[1] 彭红光[3] 李志扬[1] 

机构地区：[1]华中师范大学物理科学与技术学院,湖北武汉430079 [2]湘南学院电子信息与电气工程学院,湖南郴州423000 [3]湘潭大学物理与光电工程学院,湖南湘潭411105

出　　处：《安徽工业大学学报（自然科学版）》2015年第4期372-377,共6页Journal of Anhui University of Technology（Natural Science）

基　　金：湖南省高校科技创新团队资助项目(湘教通2014207);湖南省自然科学基金项目(13JJ3121);湖南省教育厅资助项目(13C879)

摘　　要：在场景中添加与覆盖区域颜色相近的物体,改变光照方向和强度,构成1幅新的图像。为从新图像中检测出新添加的物体(变动物体),且不受光照改变的影响,采用多尺度结构张量的方法提取图像纹理特征,采用归一化RGB色彩模型提取图像色彩特征,2种特征通过自适应权重系数结合,由此检测变动物体。采用光照变化条件下颜色纹理特征不同的图像进行对比实验,结果表明:在光照缓慢变化时,采用多尺度结构张量的方法和归一化RGB色彩模型可抑制光照变化的影响;两者结合能有效检测出变动物体。Add a new object with similar color of the covered area to a scene, built another picture with changed intensity and direction of illumination light. In order to detect the added object under the interference of changed light, firstly the textures of the pictures were extracted using multi-scale structure tensor method, secondly the col- ors of the pictures were transformed by normalized RGB model, and finally the changing of textures and colors were integrated with self-adapted weighting coefficients. Using pictures captured under varying illuminate light with different color and texture for experiments, it is found that the multi-scale structure tensor method and nor- malized RGB model are least influenced by slow varying light, when incorporated they provide an efficient means for change obiect detection.

关 键 词：变动物体检测 光照条件 多尺度小波 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]