基于单目视觉里程计的移动机器人自定位方法  

Self-Positioning Method of Mobile Robot Based on Monocular Visual Odometer

作　　者：雷越 邓斌[1] 何沛恒 徐新 左荣 LEI Yue;DENG Bin;HE Peiheng;XU Xin;ZUO Rong(Engineering Research Center of Advanced Drive Energy Saving Technologies,Ministry of Education,Southwest Jiaotong University,Chengdu 610031,China)

机构地区：[1]西南交通大学先进驱动节能技术教育部工程研究中心,四川成都610031

出　　处：《机械》2021年第4期55-61,共7页Machinery

摘　　要：为解决在列车底部直线移动的机器人对风管进行识别定位摘解时需到达两节车厢之间的自定位问题,提出一种基于单目视觉里程计的自定位方法。该方法根据单目相机在列车底部竖直向上拍摄时图像颜色较深,而当相机将要到达两节车厢之间时所拍摄图像中会有明显亮度变化区域的特点,选取关键帧图像,通过图像处理提取出图像中目标区域信息,根据所提出的运动估计算法估计相机自身的运动。最后,通过实验计算得到该方法的绝对误差为0.059 s,相对误差为1.16%,验证了该方法的可行性。In order to solve the self-positioning problem that the robot moving linearly at the bottom of the train is expected to reach the middle of two carriages when identifying,locating and disassembling the air duct,a self-positioning method based on monocular visual odometer is proposed.On the basis of the feature that the color of the image is darker when the monocular camera shoots vertically upwards from the bottom of the train,and when the camera is about to reach the middle of the two carriages,there will be obvious brightness changes in the image,the key frame image is selected,and the target area information in the image is extracted through image processing,and the camera's own motion is estimated according to the proposed motion estimation algorithm.It turns out that the absolute error of the method is 0.059s and the relative error is 1.16%,which verifies the feasibility of the method.

关 键 词：移动机器人 单目视觉里程计 关键帧选取 图像处理 运动估计 

分 类 号：TP24[自动化与计算机技术—检测技术与自动化装置]