中国科学院机构知识库网格
Chinese Academy of Sciences Institutional Repositories Grid
面向宇航机器人面板操作的视觉测量关键技术研究

文献类型:学位论文

作者丁纯杰
学位类别硕士
答辩日期2017-05-24
授予单位中国科学院沈阳自动化研究所
授予地点沈阳
导师郝颖明
关键词宇航机器人 双目立体视觉 目标识别 位姿解算
其他题名Study on the Key Technology of Visual Measurement for Panel Operation of Robonaut
学位专业控制工程
中文摘要宇航机器人是机器人技术、机器人视觉技术以及航天技术的结合,也是未来空间技术的重要发展方向之一,将会协助甚至代替人类宇航员进入空间站执行繁重的在轨任务。其中视觉测量技术是其准确获取操作对象的位置和姿态的重要保证,是宇航机器人能顺利且高效执行任务的重要前提,因此研究宇航机器人的视觉测量技术具有非常重要的应用价值。本文以宇航机器人为应用背景,针对其面板操作常用的两类目标:按钮类目标和扶手类目标,开展双目立体视觉测量关键技术研究。 按钮操作类目标属于二维目标,需要测量的是目标中心的三维位置。本文提出了一种适合多种按钮目标的三维位置测量方法,包括按钮目标识别和三维位置求解两个步骤。在目标识别阶段,提出了多特征融合的按钮目标识别方法,通过选取多种目标特征,采用特征融合策略,实现了目标的准确识别;在三维位置求解阶段,提出了三维位置优化求解方法,利用多目标间位置约束提供的冗余信息进行位置优化求解,提高了位置测量的精度。 扶手类操作目标属于三维目标,需要测量的是目标的位置和姿态。本文提出了一种扶手类目标的6维位姿测量方法,包括目标区域确定、立体匹配和位姿求解三个步骤。在目标区域确定阶段,通过多特征融合的识别方法获得目标区域;在立体匹配阶段,提出了一种基于目标边缘信息的灰度相关性匹配方法,结合目标边缘特征与外极线约束,大大减少了立体匹配的搜索空间,提高了匹配速度;在位姿求解阶段,提出了一种自适应摄像机成像角度的特征点匹配方法,利用匹配准确的特征点采用向量法完成了目标的位姿求解。 基于上述测量方法,本文建立了宇航机器人面板操作视觉测量试验系统,并开发了视觉测量软件,在多个实验场景下对其进行了实验验证,结果表明本文算法具有较好的目标识别率及鲁棒性,位姿测量精度满足工程应用要求。其中按钮类操作目标测量算法已经应用到实际工程项目中,并且成功完成了地面演示实验。
英文摘要“Robonaut”, combined with robot technology, robot vision technology and space technology, is one of the important development directions of future space technology. It will replace human astronauts into the space station to perform heavy on-orbit missions. The visual measurement technology is an important guarantee for robonaut to access to the position and attitude of the operation object accurately, and is one of the prerequisites for robonaut to smooth and efficient execution of the task. Therefore, it is very important to study the visual measurement technology of robonaut. In this paper, based on the application of the robonaut, the key technologies of binocular stereo vision measurement are studied for the two kinds of common targets of the panel operation, such as the button target and the armrest target. The button operation class object belongs to the two-dimensional target, and the three-dimensional position of the target center needs to be measured. In this paper, a three-dimensional position measurement method is proposed, which includes multi-features fusion button target recognition method and three-dimensional position optimization method based on multi-objective position constraint. In the target recognition stage, a multi-features fusion button target recognition method is proposed. By using the feature fusion strategy, the target recognition is realized by selecting various target features. In the three-dimensional position solving stage, a three-dimensional position optimization method is proposed. Multi-objective location constraints provide redundant information for position optimization, and improve the accuracy of position measurement. The armrest operation target belongs to the 3D target, and the position and attitude of the target need to be measured. In this paper, a 6-dimensional pose measurement method of handrail target is proposed, which includes three steps: target region determination, stereo matching and pose solution. In the stage of target region determination, the target region is obtained by the method of multi-features fusion. In the stereo matching stage, a gray correlation method based on target edge information is proposed, which is combined with the target edge feature and the epipolar constraint, improves the speed of stereo matching processing. In solving position stage, a feature point matching method of adaptive camera imaging angle is proposed. The vector method is used to solve the target pose of the three-dimensional target using the matched feature point. Based on the above measurement method, this paper establishes the visual measurement test system of aerospace robot panel operation, and develops the visual measurement software. The result, which is tested in multiple experimental scenarios, shows that the algorithm has a grate target recognition rate and robustness, and pose measurement accuracy can meet the accuracy of engineering applications. In addition, the button class operation target measurement algorithm has been applied to the actual project, and successfully completed the ground demonstration experiment.
语种中文
产权排序1
源URL[http://ir.sia.cn/handle/173321/20543]  
专题沈阳自动化研究所_光电信息技术研究室
推荐引用方式
GB/T 7714
丁纯杰. 面向宇航机器人面板操作的视觉测量关键技术研究[D]. 沈阳. 中国科学院沈阳自动化研究所. 2017.

入库方式: OAI收割

来源:沈阳自动化研究所

浏览0
下载0
收藏0
其他版本

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。