CASIA OpenIR  > 毕业生  > 博士学位论文
基于点云数据的表面检测与目标定位关键技术研究
其他题名Research on Key Problems of Point Cloud Based Surface Inspection and Target Localization
吴倩
学位类型工学博士
导师徐德 ; 邹伟
2015-05-26
学位授予单位中国科学院大学
学位授予地点中国科学院自动化研究所
学位专业控制理论与控制工程
关键词点云数据 表面检测 路径规划 光学导航 目标定位 Point Cloud Surface Inspection Path Planning Optical Navigation Target Localization
摘要点云数据是一种表征物体三维模型的数据格式,它具有获取方便、数据结构简单、表达能力强、细节表征丰富等优点,现已成为工程应用中常用的数据格式。本文针对基于点云数据的表面检测和目标定位中的若干关键技术进行了深入研究,并面向复杂工件表面自动化检测和深空光学自主导航这两个应用领域展开相关研究工作。论文主要工作包括: (1)面向复杂工件表面检测任务,建立了基于六自由度机械臂的自动扫描系统,在分析扫描系统测量原理的基础上,建立了系统测量模型,给出了系统标定方法。首先,分析了六自由度自动扫描系统的结构;其次,分析了结构光扫描仪测量原理,总结出需要标定的扫描仪参数;然后,建立了六自由度自动扫描系统的测量模型,指出需要标定的系统参数;最后,提出了系统标定方法,并实现六自由度自动扫描系统的标定。 (2)面向复杂工件表面检测任务,提出了基于模型的扫描路径规划算法。首先,分析了扫描路径规划的约束条件;其次,分析了三角网格模型的数据结构,基于模型的数据结构提出了网格细化算法和顶点邻域搜索算法;然后,分别提出了基于区域生长的扫描视点生成算法和基于模型框架的扫描视点生成算法,通过对比实验对两个算法进行了比较分析;接着,提出了基于贪婪遗传算法的最短扫描路径生成算法,生成机械臂末端实际扫描路径;最后,通过仿真实验和实际扫描实验验证了扫描路径规划算法的有效性,其中,实际扫描实验基于六自由度扫描系统,以机翼翼面为扫描对象,用规划的扫描路径获取机翼翼面点云数据,并进行了机翼翼面检测分析。 (3)提出了一种基于点云模型的不规则目标轮廓特征库构建方法。首先,建立了视点空间采样方法,获取采样视点的位姿信息;其次,提出了基于点云模型的不规则目标二维成像轮廓生成和提取算法;然后,面向不规则目标视线矢量提取这一问题,选择合适的轮廓特征,并提出了轮廓特征提取方法;最后,用一个不规则天体和一块不规则石头的点云模型验证了轮廓特征库构建方法的有效性。 (4)面向深空自主光学导航这一应用领域,提出了基于轮廓特征的不规则目标视线矢量提取方法。首先,分析了视线矢量的计算模型,提出了视线矢量计算的关键问题在于提取目标三维中心的成像点坐标;然后;提出了基于轮廓特征库的视线矢量提取方法,基于轮廓相似性进行最佳匹配模板搜索,基于最佳匹配模板进行视线矢量计算;最后,以不规则天体的仿真图像和不规则石头的实际图像为例,验证了本文所提出的视线矢量提取方法的有效性,并通过对比实验验证了本文所提出的方法相较于常用的轮廓质心法对不规则目标的视线矢量提取具有更高的精度。 本文以复杂工件表面自动化检测和深空光学自主导航这两个应用为切入点,深入研究了基于点云数据的表面检测和目标定位的若干关键技术,具有重要的理论意义和现实意义。
其他摘要Point cloud is a kind of digital media to represent the three dimensional model of an object, which has become a common data structure used in engineering application with the advantages of convenient acquisition, simplicity of data structure, strong expression ability, and rich detail description. This thesis focuses on the key technologies of point cloud based surface inspection and target localization. The application fields include automatic surface inspection of complex workpiece and optical deep space navigation. The main contributions are addressed as follows. (1) A 6 degree of freedom robot based scanning system is built for workpiece surface inspection. The measurement principle, measurement model, and calibration method of the scanning system are also analyzed. Firstly, the structure of the scanning system is introduced. Secondly, the measurement principle of the structured light scanner is described, and the scanner parameters which are required to be calibrated are pointed out. Thirdly, the measurement model of the scanning system is built, and the system parameters which are required to be calibrated are also pointed out. Fourthly, the calibration method for the scanning system is proposed and implemented. (2) A model-based scanning path planning method is proposed for workpiece surface inspection. Firstly, the scanning constraints used in path planning are analyzed. Secondly, the data structure of a triangular mesh model is introduced. And a mesh refinement algorithm and a neighborhood searching algorithm are proposed based on the data structure. Thirdly, a region growth based viewpoint generation method and a model framework based viewpoint generation method are proposed, respectively. Comparison experiments are presented between the two methods. Fourthly, a greed genetic algorithm based shortest scanning path generation method is proposed to generate practical scanning path of the robot end. Lastly, the path planning method is tested to verify its effectiveness through simulated and practical experiments. An airplane wing surface is taken as an example for the practical experiments based on the automatic scanning system. The surface inspection results of the airplane wing are presented. (3) A point cloud based contour feature library building method is proposed for irregular objects. Firstly, a viewpoint sampling method is proposed to obtain the pose and position of viewpoints. Secondly, a point cloud based contour generation and extrac...
其他标识符201218014628021
语种中文
文献类型学位论文
条目标识符http://ir.ia.ac.cn/handle/173211/6690
专题毕业生_博士学位论文
推荐引用方式
GB/T 7714
吴倩. 基于点云数据的表面检测与目标定位关键技术研究[D]. 中国科学院自动化研究所. 中国科学院大学,2015.
条目包含的文件
文件名称/大小 文献类型 版本类型 开放类型 使用许可
CASIA_20121801462802(4186KB) 暂不开放CC BY-NC-SA请求全文
个性服务
推荐该条目
保存到收藏夹
查看访问统计
导出为Endnote文件
谷歌学术
谷歌学术中相似的文章
[吴倩]的文章
百度学术
百度学术中相似的文章
[吴倩]的文章
必应学术
必应学术中相似的文章
[吴倩]的文章
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

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。