项目名称: 基于3D视频的运动分割与3D运动估计
项目编号: No.60872069
项目类型: 面上项目
立项/批准年度: 2009
项目学科: 金属学与金属工艺
项目作者: 于慧敏
作者单位: 浙江大学
项目金额: 30万元
中文摘要: 本项目将借助于实时三维获取技术,研究利用深度信息加二维图像(2D plus Depth)的3D视频实现运动分割和3D运动估计的原理和方法。通过该项目的研究,探索一种新的原理和方法,解决运动分割和运动估计的鲁棒性问题。 本项目研究的主要指导思想是将有着相互联系的运动分割和3D运动估计作为一个整体的求解问题。在研究方法上,利用了变分法、活动轮廓和水平集等现代数学方法,研究一种新的建模思路和方案,巧妙地将运动分割和3D运动参数等求解问题结合在一起;利用深度信息,研究如何建立精确地运动参数模型,能够处理复杂运动模式,提高运动分析的鲁棒性;研究一种新的时空域处理模型,使运动分割包含了运动跟踪,能够解决相同目标间的配对问题,同时也把更多的数据用于处理,提高了算法的鲁棒性。 所研究方法要求能够同步实现运动分割和3D运动估计,并对背景运动进行补偿,以获得真实的三维运动分析。
中文关键词: 运动分割;3D运动估计;3D视频;变分法;水平集
英文摘要: This project researches on the principles of motion segmentation and 3D motion estimation from a 3D video (2D video plus depth) obtained by the real time 3D capturing. The goal of this project is to explore a novel method to resolve the robust problems of motion segmentation and motion estimation. The key idea of this is that the motion segmentation and the 3D motion estimation are interdependent and should be considered one problem to be resolved. In study, the modern mathematics of variational methods, active contour and level set will be used to explore the novel modeling method and scheme to subtly bind the motion segmentation and the 3D motion estimation together, the accurate parametric motion model to able handle complicated motion pattern and improve the robustness of motion analysis using depth information, and the novel spatiotemporal processing method to perform tracking or matching of the same moving objects by motion segmentation and to use much more data to improve the robustness of the algorithm. The methods studied in this project are desired that the motion segmentation and the 3D motion estimation are carried out synchronously with concurrent background motion compensation for actual 3D motion analysis.
英文关键词: motion segmentation;3D motion estimation; 3D video;variational methods;level set