基于全景视觉的移动机器人的运动目标检测

在动态背景下的运动目标检测中,由于目标和背景两者都是各自独立运动的,在提取前景运动目标时需要考虑由移动机器人自身运动引起的背景变化。仿射变换是一种广泛用于估计图像间背景变换的方法。然而,在移动机器人上使用全方位视觉传感器(ODVS)时,由于全方位图像的扭曲变形会 造成图像中背景运动不一致,无法通过单一的仿射变换描述全方位图像上的背景运动。将图像划分为网格窗口,然后对每个窗口分别进行仿射变换,从背景变换补偿帧差中得到运动目标的区域。最后,根据ODVS的成像特性,通过视觉方法解析出运动障碍物的距离和方位信息。实验结果表明,提出的方法能准确检测出移动机器人360°范围内的运动障碍物,并实现运动障碍物的精确定位,有效地提高了移动机器人的实时避障能力。

Moving Object Detection in Omnidirectional Vision-based Mobile Robot

For moving object detection in dynamic background,the movement of moving object needs to be extracted by considering the background which has also changed by the ego-motion of mobile robot.Affine transformation is widely used to estimate the background transformation between images.However,using omnidirectional vision sensors (ODVS) in mobile robot will cause inconsistencies in the background motion image due to distortion of the omnidirectional image.So only one affine transformation model can not represent the whole background changes.In this paper,the image was divided into grid windows and the area of moving objects were obtained from the background transformation-compensated frame difference using every affine transformation for each window.Finally,according to the imaging characteristics of ODVS,we obtained the distance and orientation information of moving obstacle by visual method.The results demonstrate that the proposed method is very efficient in moving object detection.It can also realize the precise localization of moving obstacles and greatly improve the ability of abstacle avoidance of mobile robot.