基于轮廓特征点的双目视觉体积估测模型  

Volume Estimation Model Based on Contour's Feature Points and Binocular Vision

作　　者：胡亚楠 潘焜 王铁铮[1] 喻晓 HU Ya-nan;PAN Kun;WANG Tie-zheng;YU Xiao(Material Branch of State Grid Beijing Electric Power Company,Beijing 100053)

机构地区：[1]国网北京市电力公司物资分公司,北京100053

出　　处：《现代计算机》2020年第11期45-49,共5页Modern Computer

摘　　要：面对仓储管理设计一种新的双目视觉物体体积估测模型。先对使用的双目摄像头进行标定。校正双目摄像头采集的图像,立体匹配并计算视差,通过边缘检测和三维重建得出物体规范化后的轮廓特征点三维世界坐标按长方体体积计算方式完成体积的估测。在实验条件下对若干物体体积进行估测,误差率为20%,验证估测模型的可行性。另外探究模型运行中花费最多时间的模块和取得良好估测结果时摄像头的工作范围两个课题。Proposes a new volume estimation model based on binocular vision as for warehousing management.The model calibrates binocular camera firstly,then through stereo matching calculates the disparity in two rectified pictures got by camera and through edge detection and 3D reconstruction obtains contour’s regularized feature points’3D coordinates to estimate the object’s volume.Some experimental results show that error rate of the model is 20%,which validates its feasibility.Some subjects,such as the module which costs the most time in the model and camera’s proper working position to get a good result,was also explored.

关 键 词：双目视觉 体积估测 轮廓检测 三维重建 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] F274[自动化与计算机技术—计算机科学与技术]