基于区域增长的图像跟踪算法的研究  

Study on image tracking base on region growing

作　　者：薛雪[1] 刘泽平[1] 丁艳[1] 

机构地区：[1]北京理工大学机电工程学院,北京100081

出　　处：《光学技术》2005年第1期152-154,共3页Optical Technique

摘　　要：为了提高序列图像的跟踪精度,提出了基于区域增长的特征量提取方法。这种方法可捕获图像中所有单连通域,并能准确地提取其特征量,然后依据所提取的特征量设计分类器,实现对图像中所有单连通域进行分类识别并加以跟踪的目的。此算法有效地解决了一般传统识别算法难以区分目标和其近邻区域的干扰,而导致跟踪目标特征量提取不准确的问题。在目标特征提取识别算法的基础上,还提出了阈值预测分割算法,并应用在序列图像的跟踪中,取得了较好效果。In order to improve the precision of image tracking, a new recognition arithmetic used in region growing is proposed. The method can capture the single connected component in image and extract it's feature. According to the feature, the method of categories to recognize entire single connected components is design. The recognition arithmetic solves inexact extracted feature, which is brought on by disturbance of background image near the object in ordinary recognition arithmetic. Base on the recognition arithmetic, a segmentation method by forecasting threshold is proposed. Using this segmentation method in image tracking, satisfying effect is gotten.

关 键 词：连通域 区域增长 识别算法 序列图像 分类器 图像跟踪 分割算法 目标特征提取 跟踪精度 捕获 

分 类 号：TH21[机械工程—机械制造及自动化] TP391[自动化与计算机技术—计算机应用技术]