粗糙集在智能空瓶检测感兴趣区域提取中的应用  

Application of Rough Sets to Extract Region of Interest in Intelligent Empty Bottle Inspection

作　　者：张东波[1] 王耀南[1] 段峰[1] 刘焕军[1] 

机构地区：[1]湖南大学电气与信息工程学院

出　　处：《系统仿真学报》2007年第5期1021-1025,共5页Journal of System Simulation

基　　金：国家自然科学基金(60375001);高等学校博士点基金(20030532004);湖南省教育厅资助科研项目(05C093)

摘　　要：为实现空瓶检测中感兴趣区域ROI的自动精确标定,介绍了一种基于粗糙集不可分辨关系划分及粗近似进行ROI区域提取的新方法。首先,基于先验知识描述,确定粗略ROI区域,然后,提取和ROI区域标定有关的底层图像特征如灰度、边缘、位置等,在对特征属性离散化后,构造出反映分类关系的信息表,并依据不可分辨关系划分获得基本像元区域,最后,以初始ROI区域的上近似作为最终提取的ROI区域。在瓶身及瓶口的ROI区域提取实验中,该方法可以获得比人工标定更为精细的ROI区域,有利于提高后续检测过程中的检测精度。A new method which could extract region of interest（ROI） was introduced. Based on partitions of indiscernibility relations and rough approximation, ROI could be extracted automatically and precisely. First, based on prior knowledge, a rough ROI was determined, Next, by extracting the low-level features such as intensity, edge, location and so on, which are correlated with marking ROI, an information table reflecting the relation of classification could be constructed and basic regions were built after discretization of the attributes. Finally, final ROI could be represented by the upper approximation of original rough ROI. In the experiment of bottle body and bottle mouth for ROI extraction, this method acquired more precise ROI regions than manual marking, which benefits later inspection of the bottle.

关 键 词：粗糙集 空瓶检测 感兴趣区域ROI 粗近似 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]