基于Markowitz投资组合模型的移动机器人目标提取与识别

Object extraction and recognition for mobile robot based on Markowitz portfolio model

移动机器人足球比赛中,准确识别足球并确定其位置是比赛胜负的关键。在理想情况下,比赛场地颜色信息固定,可以通过彩色分割来识别,但是若存在环境光照较弱或不均的影响时,很难明确划分不同颜色区域之间的界限,使得对目标的准确识别受到影响。提出一种全新的彩色分割方法,通过适当选择几个不同的颜色分量,应用Markowitz投资组合模型对各个颜色分量求取最优权值,进而将最优权值作用于彩色图像,可得到不同区域间对比度增强的灰度图像,分割出所求目标。实验表明,该方法很好地弱化了光照变化带来的影响。

In mobile soccer games,high-accurately identify the soccer and determine its position are key to the final result.In ideal cases,the color information of the court is stable,and the soccer can be distinguished from surrounding objects based on the color image segmentation.However,when under the environment with weak or uneven illumination,the boundary between the different colors＇ domain is hard to plot definitely,which could affect the final recognition of the objectives.In this paper,a novel method for color image segmentation was proposed.According to this method,the properly selection of color component could be of great importance by invoking Markowitz portfolio model to obtain the optimal weights of the color components.Moreover,the achieved optimal weights were provided for the color images＇ processing,which could help to obtain the enhanced gray images within different domains and separate the object expected.The results demonstrate that the influences of illumination variation are weakened by the proposed method.