MiroSot足球机器人视觉图像信号处理算法改进研究

【作者】 方成；

【导师】 喻寿益；

【作者基本信息】 中南大学 ， 控制理论与控制工程， 2002， 硕士

【摘要】 足球机器人比赛是近年来国际上兴起的一种高科技竞赛活动。足球机器人分为两大系列：MiroSot系列和RobCup系列。足球机器人是集机器人学、智能控制、数据融合、计算机技术、无线通讯、图像处理、机械学等多种学科和技术于一体的综合性产物，它本身又是一个很典型的非线性的控制对象，为各种控制理论提供了一个非常好的实验平台。 MiroSot系列足球机器人主要由视觉子系统、决策子系统、通讯系统、机器人小车系统四部分组成。本文阐述MiroSot系列足球机器人视觉子系统图像信号处理算法的改进。文中首先介绍了数字图像处理方面的一些基础知识和MiroSot足球机器人视觉子系统的一些实现技术，然后分析了从东北大学引进的足球机器人视觉子系统存在的误识别、丢失目标以及识别出的目标位置精度不高这三个方面的问题，针对这些问题提出了四种改造方法。这些方法分别是：建立自适应多值化颜色查找表解决误识别问题；通过高阶插值算法改造目标图像放大方式提高色标采集准确率从而减小丢失目标的可能性；通过位置补偿的方法来矫正由于摄像机镜头的光学作用引起的图像形变；采用一阶矩目标中心定位法提高目标中心点的计算精度。 文中提出了一种根据彩色图像ROB分量，分割目标与背景的有效方法，实验表明这种方法能将目标与背景完好地分离出来，计算速度快，大大地减少了目标错误识别率，提高了定位精度，具有较好的实际应用价值。更多还原

【Abstract】 Robot soccer match is a kind of high technology match which has arisen around the world in recent years. Now, robot soccer is divided in two kinds. The first one is MiroSot series and the second is RobCup series. Robot soccer, referring to the fields of robotics -> intelligent control, data fusion, computer technology-, wireless communication, image processing, mechanics and so on, is the outcome of all these scientific technology. It is also a typical nonlinear control object and is a good test bed for many kinds of control theory.The whole system of MiroSot robot consists of four subsystems: robot vision, decision-making and communication subsystem. The paper is about amelioration of image processing algorithm of robot soccer vision system in MiroSot soccer robot, it firstly elaborates some basic knowledge on image processing then analysis three problems in the robot soccer vision system which was bought from North-east University. The first problem is fault recognition; the second is about losing target; the last one is about low accuracy of target position. The paper presents four ways to solve the problems. The first way is building multi-color table to solve fault recognition; the second method is changing the way of enlarging target map to decrease the possibility of losing target; the third technique is using position compensation to rectify distortion of image caused by the lens of vidicon; the last method is using the mathematical characteristic of quadrature to improve the accuracy of target position.This paper presents a valid method to separate the target from the background according to RGB of color images, this method which is proved by experiment, is powerful in image segment. It has good value in practice for its high speed, low target-finding fault and accuracy position.更多还原