项目名称: 基于解析模型的不同场景下行人异常行为的识别与跟踪
项目编号: No.61202327
项目类型: 青年科学基金项目
立项/批准年度: 2013
项目学科: 计算机科学学科
项目作者: 薄一航
作者单位: 北京电影学院
项目金额: 23万元
中文摘要: 智能视频监控利用计算机视觉技术在无需任何人工参与的情况下对视频数据进行处理、分析和理解,进而分析与判断目标的行为,能够在发生异常情况时实时提供有用信息并发出相应的警报,帮助人们预防和处理各种危机。行人的异常行为预测与跟踪是目前最具应用与经济价值的研究课题之一,然而其仍面临着许多挑战,离实际应用还有一段距离。本项目拟针对不同场景下同一行人目标的识别、其异常行为的预测及跟踪等问题展开研究,其主要工作包括:(1) 基于解析模型的不同场景下同一行人目标的识别,(2) 基于解析模型的行人异常行为预测, (3) 基于解析模型与3D姿势估计的行人的跟踪。本项目提出的新模型简单易行,且具有坚实的研究基础,对于智能监控技术的发展与进步具有特别重要的意义与巨大的应用价值。
中文关键词: 视频分割;行为识别;显著目标;;
英文摘要: Intelligent Video Surveillance can process, analyze and understand video data using the technology of computer vision without human's participantion. It locates, recognizes and tracks the changes of the scene in image sequence and determines the behavior of the object, which can alarm the accident under abnormal cases and help human to predict kinds of dangers. Pedestrian abnormal behavior prediction and tracking is one of the most popular projects which has application and economic value. However, it also faces some challenges which is far from the practical application. This project aims to recognize the same pedestrian under different scenes, predict its abnormal behavior and track it. It will focus on three key points: (1) Parsing-based pedestrian recognition under different scenes, (2) Parsing-based pedestrian abnormal behavior prediction, (3) Parsing and 3D pose estimation based pedestrian tracking. The model that will be proposed in this project which has strong base will be easy to achieve and it is significance to the development of intelligent video surveillance.
英文关键词: Video Segmentation;Action Recognition;Salient Object;;