基于深度图像的室内机器人避障系统  

On the obstacle avoidance system for mobile robots based on depth image in indoor condition

作　　者：马利红[1] 范晨光[1] 王书平[2] 周武[2] 

机构地区：[1]浙江师范大学数理与信息工程学院,浙江金华321004 [2]浙江师范大学工学院,浙江金华321004

出　　处：《浙江师范大学学报（自然科学版）》2016年第3期276-282,共7页Journal of Zhejiang Normal University:Natural Sciences

基　　金：国家自然科学基金青年基金项目资助(51405450);2014年度国家级大学生创新创业训练计划项目(201410345009)

摘　　要：为了增强室内移动机器人的障碍物检测能力和降低系统成本,提出了一种基于深度图像的单一Kinect传感器的室内移动机器人避障系统.该系统将Kinect传感器竖直放置,并对深度图像进行了几何变换,提供了更加适合探测地面和空中障碍物的视角.对深度图像进行中值滤波以消除图像噪声,并采用统计平均的背景减除法去除无障碍背景,提取障碍物的深度图像,然后对障碍物的大小、位置与数量进行检测,最后根据检测的结果,并利用人工视场与Kinect摄像头对位置环境局部最小值场景的检测相结合进行避障路径的选择.结果表明:局部最小值场景检测方法能够高效检测出环境中存在局部最小值的缺陷,沿墙行走模式也大大减少无效路径,解决了狭窄通道内震荡的问题,有效地实现了室内机器人在动态环境下的避障.An obstacle avoidance system for mobile robots was designed to improve obstacle detection and re- duce costs used a single sensor Kinect based on the depth images in indoors condition. The Kinect sensor was placed vertically, and the depth image was geometry transformed in this system, which provided a more suit- able field to detect obstacles of ground and air. The depth image by median filtering eliminated image noise ef- ficiently, and through the statistical average background subtraction division removing barrier-free background to extract the obstacles of depth image, and then detected the size, location and quantities of the obstacles, according to the test results using the combination of artificial field and Kinect device camera to local minimum position environment scene detection for path choice avoiding walls. The experimental result showed that the local minimum value of scene detection method could efficiently detected the local minimum defects in the environment scene, the invalid path of walking along the wall model was greatly reduced, the problem of the narrow channels of concussion was solved, interior obstacle avoidance of robot in dynamic environment was effectively implemented.

关 键 词：深度图像 室内机器人 KINECT 避障 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]