Ma Wei, Ma Shoulei, Bi Jianjian, Yin Chen-bo, Cao Donghui, Yuan Hongfu
{"title":"Three-dimensional Reconstruction of Working Environment in Remote Control Excavator","authors":"Ma Wei, Ma Shoulei, Bi Jianjian, Yin Chen-bo, Cao Donghui, Yuan Hongfu","doi":"10.1109/ICMCCE51767.2020.00029","DOIUrl":null,"url":null,"abstract":"Remote control excavator solves the personal safety problem of operators in dangerous conditions, but the video surveillance of remote control excavators can't sense the stereo distance of the target. To solve the problem, the binocular vision method is used to reconstruct the 3D operating scene of the excavator, and help the drivers to effectively sense the spatial position of the target and bucket to improve the efficiency of remote control operation. Firstly, the noise generated by binocular camera during data collection was removed by guided filtering. And then, left and right images were matched by features based on Sum of Absolute Differences (SAD). Then, the generated disparity value was optimized according to the Winner Takes All(WTA) strategy. Finally, the disparity map of the working scene was obtained and the 3D reconstruction of the scene was completed. The experimental results show that the binocular 3D reconstruction system used in this paper can realize the 3D reconstruction of different scenes indoors and outdoors through the cooperation of software and hardware.","PeriodicalId":6712,"journal":{"name":"2020 5th International Conference on Mechanical, Control and Computer Engineering (ICMCCE)","volume":"136 1","pages":"98-103"},"PeriodicalIF":0.0000,"publicationDate":"2020-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2020 5th International Conference on Mechanical, Control and Computer Engineering (ICMCCE)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ICMCCE51767.2020.00029","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
Remote control excavator solves the personal safety problem of operators in dangerous conditions, but the video surveillance of remote control excavators can't sense the stereo distance of the target. To solve the problem, the binocular vision method is used to reconstruct the 3D operating scene of the excavator, and help the drivers to effectively sense the spatial position of the target and bucket to improve the efficiency of remote control operation. Firstly, the noise generated by binocular camera during data collection was removed by guided filtering. And then, left and right images were matched by features based on Sum of Absolute Differences (SAD). Then, the generated disparity value was optimized according to the Winner Takes All(WTA) strategy. Finally, the disparity map of the working scene was obtained and the 3D reconstruction of the scene was completed. The experimental results show that the binocular 3D reconstruction system used in this paper can realize the 3D reconstruction of different scenes indoors and outdoors through the cooperation of software and hardware.
遥控挖掘机解决了危险工况下操作人员的人身安全问题,但遥控挖掘机的视频监控无法感知目标的立体距离。为解决这一问题,采用双目视觉方法重构挖掘机的三维操作场景,帮助驾驶员有效感知目标和铲斗的空间位置,提高遥控操作效率。首先,采用引导滤波方法去除双目摄像机采集数据过程中产生的噪声;然后,基于绝对差和(Sum of Absolute difference, SAD)的特征对左右图像进行匹配。然后,根据赢家通吃(WTA)策略对生成的视差值进行优化。最后,获得工作场景的视差图,完成场景的三维重建。实验结果表明,本文所采用的双目三维重建系统通过软硬件的配合,可以实现室内外不同场景的三维重建。