基于视觉传感器的障碍物检测

摘要

针对移动机器人双目视觉障碍物检测的实时性和准确性两大难题，提出了一种适用于室内复杂背景的障碍物检测方法。通过Retinex图像增强操作消除环境光照不均匀的影响；运用扫描线种子填充算法分割图像中的路面、背景和障碍物；通过二值化处理、边缘检测提取障碍物的轮廓信息和位置信息。在AS-RE轮式机器人平台上进行实验，实验结果表明，机器人能够在室内环境中稳定地实现自主避障功能，验证了提出的双目视觉障碍物检测算法的可行性。%To solve real-time and accuracy problems of mobile robot obstacle detection based on binocular vision, a new obstacle detection method for complicated indoor background is proposed. The influence of non-homogeneous illumination is removed via Retinex image enhancement operation; The seed filling is used to implement segmentation of road face, background and obstacle;The contour and position information of obstacle is obtained after binaryzation, and edge detection. The test result on AS-RE robot shows that the new obstacle detection method is viable, and the automatic obstacle avoiding function is well realized stably.