PRoCeSS: Pose Recovery in Context Specific Scenarios

过程：特定场景中的姿势恢复

• 批准号: EP/E033288/1
• 负责人: Dimitrios Makris
• 金额: $ 13.75万
• 依托单位: Kingston University
• 依托单位国家: 英国
• 项目类别: Research Grant
• 财政年份: 2007
• 资助国家: 英国
• 起止时间: 2007 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=EP%2FE033288%2F1
• 关键词: PRoCeSS; Pose; Recovery; Context; Specific

The aim of PRoCeSS is to advance the state-of-the-art in Pose Recovery (i.e. determining the position and orientation of articulated parts of the human body) to the level that can operate reliably outside specialised labs, in real-life scenarios.The PRoCeSS system will be useful in applications such as the coverage and the analysis of sporting events. Athletes' movements will be captured in a 3D representation and analysed. TV will be able to virtually replay sporting events, which will be constructed automatically. Human-computer interfaces based on 3D body motion will provide people with a physical way to interact with machines. Also, visual surveillance will acquire another tool for analysing human behaviour.PRoCeSS will have to challenge real-life restrictions such as cluttered scenes, single cameras or multiple cameras that are well separated from each other. It will attempt to address these issues by exploiting the context of specific scenarios (e.g walking, chatting, opening the door in surveillance, running, kicking the ball in football, serving the ball in tennis) to increase the feasibility of real life Pose Recovery. It will also exploit the link between certain activities and areas of the scene (e.g. sitting down on seat , opening door at the exit, typing keyboard at computer desk ), or the specific sequencing of actions in particular scenarios (e.g. the scenario penalty , consists of the following activities of the player: placing the ball , walking backwards , running forwards and kicking the ball )

PRoCeSS的目标是将姿态恢复（即确定人体关节部分的位置和方向）的最新技术提升到可以在专业实验室之外的真实场景中可靠操作的水平。PRoCeSS系统将在体育赛事的报道和分析等应用中发挥重要作用。运动员的动作将以3D形式被捕获并进行分析。电视将能够虚拟重播体育赛事，这将是自动构建的。基于三维身体运动的人机界面将为人们提供一种与机器交互的物理方式。此外，视觉监视将获得另一种分析人类行为的工具。过程将不得不挑战现实生活中的限制，如杂乱的场景，单个摄像机或多个摄像机彼此分开。它将尝试通过利用特定场景（例如走路、聊天、在监视中开门、跑步、踢足球、打网球）来解决这些问题，以增加现实生活中姿势恢复的可行性。它还将利用特定活动和场景区域之间的联系（例如，坐在座位上，在出口打开门，在电脑桌旁输入键盘），或特定场景中的特定动作顺序（例如，场景惩罚，由以下玩家活动组成：放置球，向后走，向前跑和踢球）。

项目成果

Human pose tracking in low dimensional space enhanced by limb correction

通过肢体校正增强低维空间中的人体姿势跟踪

• DOI: 10.1109/icip.2011.6116100
• 发表时间: 2011
• 作者: Moutzouris A

Human pose tracking by Hierarchical Manifold Searching

通过分层流形搜索进行人体姿势跟踪

• 发表时间: 2012
• 期刊: Proceedings - International Conference on Pattern Recognition
• 作者: Moutzouris A.

Multimedia Analysis, Processing and Communications

多媒体分析、处理和通信

• DOI: 10.1007/978-3-642-19551-8_14
• 发表时间: 2011
• 作者: Nebel J

Temporal Extension of Laplacian Eigenmaps for Unsupervised Dimensionality Reduction of Time Series

• DOI: 10.1109/icpr.2010.48
• 发表时间: 2010-08
• 期刊: 2010 20th International Conference on Pattern Recognition
• 作者: Michal Lewandowski;J. M. D. Rincón;D. Makris;Jean-Christophe Nebel