嵌入式单目机器人视觉检测及避障系统设计

Design of Embedded on—Monocular Vision Detection and Obstacle Avoidance system of Robot [HS)]

摘要:

为实现机器人在移动中快速检测和避开障碍物功能,提出一种在室内以单目视觉为基础的嵌入式机器人检测障碍和避障的设计方案；通过对HSI颜色空间的H空间进行分块,获取分块区的颜色矩特征进行比较,能快速检测障碍物；通过几何法建立机器人坐标和图像坐标的关系,计算障碍物实际距离,准确定位障碍物；通过输入障碍物距离和偏离前进方向的角度到模糊控制器,获得输出转动角度,并对输出的角度进行调节；实验结果表明该系统具有实时性和可行性,实现了利用图像分块的信息快速检测定位障碍物,并通过模糊控制器完成了避障要求。

Abstract:

To realize the function of obstacle detection and avoidance on the robot moving fastly,embedded system design scheme of obstacle localization and avoidance based on monocular vision robot is proposed. With H space of the HSI space divided into three blocks,color moment features of these blocks are compared to detect obstacles. In order to locate obstacles correctly and measure distance between obstacles and the robot,a relationship between the robot coordinates and the vision coordinates is deduced by geometrical reasoning. To quickly avoid obstacles,the output rotation angle is obtained by fuzzy controller which can receive the obstacle distance and deviation angle to the original direction. The experimental results show that the system is real-time and feasible,and this system realizes the fast detection,localization with the image information and meets avoidance of obstacle through the fuzzy controller.