面向踩踏事故防范的运动人群无人机协同航拍及风险演化态势计算

项目名称： 面向踩踏事故防范的运动人群无人机协同航拍及风险演化态势计算

项目编号： No.61502433

项目类型： 青年科学基金项目

立项/批准年度： 2016

项目学科： 计算机科学学科

项目作者： 吕培

作者单位： 郑州大学

项目金额： 20万元

中文摘要： 人群踩踏是大型公共活动的主要事故类型之一，严重威胁人民群众的生命安全。由于人群运动的范围很广且形式复杂，使得踩踏事故的防范充满挑战。本课题以无人机为载体，结合多个不同学科领域的基础理论，探索现有工作尚未涉及的科学难点问题的解决方法，内容包括：(1)拟采用基于多目视频的实时人群跟踪技术和基于运动轨迹的人群态势计算技术，实现大规模人群运动的实时分析与态势计算；(2)拟采用基于兴趣区域的多相机配置技术和面向动态场景的实时相机跟踪技术，结合人群态势分析，实现面向大规模人群运动的相机控制；(3)拟采用基于融合信息的无人机自动导航技术，结合多相机动态配置和动态场景实时相机跟踪，实现大规模人群运动引导下的无人机自动导航；(4)拟采用基于能量场的踩踏行为检测技术，结合多目实时人群跟踪，构建人群踩踏行为检测模型。通过上述研究，实现面向踩踏事故防范的运动人群无人机协同航拍及踩踏风险演化态势计算，并对相关核心技术进行系统集成验证，以期为国家公共安全保障提供有益探索和技术支持。

中文关键词： 人群跟踪；踩踏预警；相机跟踪；无人机导航

英文摘要： The crowd stampede is one of the main types of accidents in the large-scale public events and threats the safety of people's life seriously. The range and form of crowd movement is very wide and complex, which makes the stampede accident prevention extraordinarily challenging. With UAV as the carrier and advantages of basic theoretical disciplines from different research fields, this proposal plans to explore the algorithms and methods for those scientific problems which are not touched by existing work: (1) we propose the technology of real-time crowd tracking based on multi-view and the technology of trajectory-based trend computation, to achieve the goal of real-time motion analysis and trend computation for the large-scale crowd; (2) we employ the technology of multi-camera configuration based on region of interest and the technology of real-time camera tracking for dynamic scene, with crowd trend analysis, to realize the camera control for the large-scale crowd movement; (3) we utilize the technology of UAV automatic navigation technology based on information fusion, with dynamic multi-camera configuration and real-time camera tracking for dynamic scene, to realize the UAV automatic navigation guided by the large-scale crowd movement; (4) we leverage the technology of stampede detection based on energy field, with real-time crowd tracking based on multi-camera, to construct the detection model for crowd stampede. Through the research above, we aim at realizing UAV cooperative aerial and risk evolution computing of moving crowd for the stampede accident prevention, and verifying related core technologies in a complete system, so as to provide meaningful exploration and technical support for the national public security.

英文关键词： crowd tracking;stampede warning ;camera tracking;UAV navigation