摘要
空间物体的理解与识别是计算机视觉与计算机图形领域的重要研究课题,广泛应用于机器人感知导航、人身安全和环境监控等方面。以点云为物体的数据表示形式,依据空间物体的形状复杂度不同,从简单场景(单目标物体)、复杂场景(多目标场景)两个方面来重点探讨,以分割、理解及识别为主线分别针对空间点云物体的理解与识别的研究现状和进展进行了归纳和总结,并通过分类讨论和剖析现有方法的优缺点,指出空间物体理解与识别方面所遇到的姿态、拓扑关联及语义等关键性问题,并提出相应的解决思路和下一步的研究重点。
Understanding and recognition of spatial objects are the important research in computer vision and computer graphics which can be widely used in robot perception navigation,personal safety and environment monitoring,etc.Based on different shape complexity of spatial objects in point clouds,this paper mainly discusses the spatial objects from simple scenario（single target object）to complex scenes（multi-target scenarios）.Taking the segmentation,understanding and recognition as mainlines,the paper summarizes the recent work and advances in understanding and recognition of the spatial objects in point clouds respectively.Also,it points out the critical issues including pose,topology and semantics encountered in the understanding and recognition of spatial objects via the discussion and analysis of the existing advantages and disadvantages,and suggests the corresponding solving-thinking ways and research priority for the future work.
出处
《西安理工大学学报》
CAS
北大核心
2014年第2期145-156,共12页
Journal of Xi'an University of Technology
基金
国家自然科学基金资助项目(61302135
61272284)
博士学科点专项科研基金资助项目(20126118120022)
中国博士后基金资助项目(2014M552469)
关键词
点云数据
空间物体
物体理解
物体识别
point cloud data
spatial objects
object understanding
object recognition
