基于烟丝颜色模型与烟支模板构造的烟支识别计数  

Cigarette recognition and counting based on construction of tobacco color model and cigarette template

作　　者：曹计昌[1] 朱圣健[1] 

机构地区：[1]华中科技大学计算机科学与技术学院,武汉430074

出　　处：《计算机辅助工程》2008年第1期56-61,共6页Computer Aided Engineering

摘　　要：为获得对烟支的高精度识别与计数,通过对烟丝颜色的实验和分析得到烟丝颜色的分布特性,在此基础上分别构造基于HSI颜色空间和RGB颜色空间的烟丝颜色模型,并利用RGB颜色空间中对应的特征向量分割烟支的内部与外部区域.根据烟支图像的特征构造烟支模板,给出基于该模板的烟支识别与计数的算法.该方法通过搜索匹配对分割出的目标区域是否为烟支进行识别与判定.实验结果表明,该算法的识别精度高,能适应一定范围的光学环境变化,鲁棒性得到增强.To recognize and count cigarettes precisely, the characteristics of tobacco color distribution is obtained through the tobacco color experiment, on which the construction of cigarette color model in HSI space and RGB space is based. The inside and outside regions of cigarette are segmented by feature vector in RGB space. A cigarette template is constructed according to the feature of cigarette image. So an algorithm on cigarette recognition and counting is proposed based on the template. The algorithm recogni- zes and judges whether the segmented target regions are cigarettes by searching and matching. The exper-imental results show that the algorithm has high recognition precision and is tolerant in a certain changing degree of optical environment, and the robustness is enhanced.

关 键 词：颜色分类 模板匹配 图像识别 形状特征 计数 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TQ533[自动化与计算机技术—计算机科学与技术]