动态多摄像头环境中拥挤多目标跟踪的联合建模与协同优化

项目名称： 动态多摄像头环境中拥挤多目标跟踪的联合建模与协同优化

项目编号： No.61305014

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 自动化技术、计算机技术

项目作者： 韩华

作者单位： 上海工程技术大学

项目金额： 25万元

中文摘要： 动态多摄像头中拥挤多目标的跟踪问题一直是智能视频跟踪领域的一个难点，联合建模和协同优化问题的研究将对其中一些难点带来突破性的研究进展。本课题旨在解决拥挤环境中目标再确认、多目标精确跟踪和多摄像头协同优化的问题。首先，对多目标空间信息和视觉信息进行联合建模，利用特征对比和几何关系检测的方式来进行目标再确认。然后，建立拥挤目标间的交互矩阵用以描述目标间的交互关系，通过查询交互矩阵的方式直接剔除虚假关联，并利用摄像头运动模型矫正目标的状态转移方程，这些处理方式可以在有效地降低计算量的同时提高多目标跟踪的准确性，达到精确跟踪。第三，多摄像头协同优化算法的提出，可以使得拥挤多目标均达到最优的跟踪和显示效果。动态多摄像头环境中拥挤多目标跟踪的联合建模与协同优化研究将为复杂跟踪问题的研究提供新的研究思路和研究成果。

中文关键词： 目标再确认；多目标跟踪；联合建模；协同关联；

英文摘要： Crowded multi-target tracking in dynamic multi-camera environments has been a difficult problem of intelligent video tracking. Jointly modeling and collaborative optimization would bring great breakthrough toward some difficulties. The purpose of this project is to solve the problems of crowded environment, such as, target re-recognition, muti-target precise tracking and multi-camera's collaborative optimization. First of all, Jointly model the spatial and visual information of multi-target, and then use the feature comparison and geometric relationship detection to re-recognize targets. Second, create a interactive matrix of crowded targets to describe the relationship between them, so we can directly excluding the false association through querying the interactive matrix. Furthermore, we use the camera motion model to correct the targets state transition equation, These treatments can improve the accuracy of crowded multi-target tracking while effectively reduce the amount of computation, and ultimately achieve precise tracking. Third, the proposition of multi-camera collaborative optimization can make crowded multi-target achieve optimal tracking and displaying. The research of jointly modeling and collaborative optimization of dynamic multi-camera environments will provide new research ideas and new achievem

英文关键词： Target Re-identification；Multi-target Tracking；Jointly Modeling；Collaborative Association；