一种基于光流计算的机器人视觉与行为模型  

The vision-action model based on optical flow processing

作　　者：梁冰[1] 洪炳熔[1] 曙光 

机构地区：[1]哈尔滨工业大学计算机科学与技术学院,哈尔滨150001

出　　处：《宇航学报》2003年第5期463-467,共5页Journal of Astronautics

基　　金：国家863计划基金资助(项目编号:2002AA734001)

摘　　要：主动环境探测与准确的行为控制近来已成为机器人领域的重点研究课题,它直接关系到机器人对环境的适应能力,以及能否根据周围环境做出及时、准确的动作规划。该研究背景是月球探测机器人的智能导航,阐述了如何建立机器人视觉与电机行为关系的视觉与行为模型。在此模型中,机器人根据运动所引起的地面图像光流场变化来计算当前机器人运动状态。在实验室环境中,该方法具有运行速度快,工作稳定等特点。这种视觉与行为模型可以用于今后的自主机器人主动目标跟踪和动态避障等实际导航项目中,具有一定的现实意义。The environment active detecting and behavior controlling for autonomous robot become very significant research fields in recent years,which is important for robot to accommodate itself to surroundings and do some action correctly and timely according to the environment.The background of this paper was based on the project of intelligent navigation for lunar explorer robot.A VisionAction model was developed to describe the relationship between the robot vision system and motor behaviors.In this model,the behavior state can generate according to the optical flow field changes of ground while the robot moving.This method is proved to be stable and short time consuming in our lab.Moreover,this VisionAction model is useful for autonomous robot to active track object and avoid dynamic obstacle,which is meaningful in widely application.

关 键 词：光流场 自主机器人 机器人导航 

分 类 号：TP224[自动化与计算机技术—检测技术与自动化装置]