一种新的实时彩色目标分割方法

为实现对视频序列图象中彩色目标的识别与跟踪,提出了一种鲁棒的实时分割方法。它分为象素分类与区域特征提取两过程。在象素分类阶段,把每一个目标象素建模成一个二维高斯模型,定义相应的决策函数,将图象中所有象素点分类成为目标和背景,同时对高斯模型的参数进行在线动态修改,使之对变化的光照条件所引起的彩色畸变具有自适应性。在区域特征提取阶段,对象素分类所形成的二值图象实现行程编码,产生行程编码映射图,在此基础上,采用一种基于树的区域搜索算法计算出区域特征量。文中提出的方法已成功地应用于足球机器人视觉子系统中,取得了良好的效果。

A Novel Real-time Segmentation Algorithm of Color Object

In this paper,a robust real-time segmentation algorithm is introduced,which makes it possible to implement color object recognition and tracking in video sequences.This algorithm includes two component ,i.e.pixel labeling and re-gion feature extraction.First,a2D Gaussians is used to model the color objects at each pixel.A match function is define to label objects pixel and backgrounds pixel.At the same time ,a technique for dynamically updating the models to ac-commodate changes in apparent colour due to varying lighting conditions is used.Second,a run length encode(RLE)is computed for the classified image,and a tree-based region find is employed to compute region feature.The algorithm has been implied to the vision system of Soccer Robot.