一种基于漫水填充法的实时彩色目标识别方法.pdfVIP

一种基于漫水填充法的实时彩色目标识别方法 陈佳鑫，贾英民 北京航空航天大学系统与控制系，北京，100191 E-mail: drpepper1@163.com, ymjia@ 摘要：本文提出了一种基于漫水填充法的实时彩色目标识别方法。经典的漫水填充法需要人工选择初始种子点，本文首先通过 直方图阈值法实现了初始种子点的自动选取，并设置了去噪邻域窗，以防止不属于目标区域的像素点被选为初始种子点。从初 始种子点出发，使用漫水填充法，得到一组互无交集的连通区域及其外接矩形。然后，利用各连通区域外接矩形几何中心之间 的距离，对连通区域进行归类。合并属于同一类的所有区域即得到完整的目标区域。实验表明：本文提出的算法能很好地识别 颜色特征突出的目标，对光照变化具有较强的鲁棒性，能完成实时彩色目标识别任务。与使用广泛的直方图阈值分割法相比， 本文算法能更加准确地对目标进行识别。 关键词：彩色目标识别，漫水填充法，直方图阈值，连通区域归类 中图分类号：O235 文献标识码：A A Real-time Color Object Recognition Method Based On Flood Fill Algorithm CHEN Jia-xin, JIA Ying-min School of Mathematics and System Science, Beihang University, Beijing 100191, P.R.China E-mail: drpepper1@163.com, ymjia@ ABSTRACT: This paper proposes a real-time color object recognition method, based on flood fill algorithm. In the traditional flood fill algorithm, initial seed pixels should be planted manually. To avoid this, the histogram-based thresholding method is utilized. In addition, a neighborhood window denoising mechanism is presented in order to prevent pixels outside the object areas from being selected as seed points. Applying flood fill algorithm at initial seeds, a class of connected areas separated from each other, together with the corresponding enclosing rectangles, are obtained. By calculating the distance between centers of any two enclosing rectangles, previous connected areas are divided into several classes. Object areas are then segmented by just combining connected areas in the same class. Experiment results indicate that the proposed method has a good performance in real-time recognition of objects with salient color characteristics , and is robust to illumination. Compared to the widely used histogram-based threshold