**作者：*Martin Hofmann1, Daniel Wolf1,*2, Gerhard Rigoll1

作者单位：

发布时间：2013

发布期刊/会议：IEEE Conference on Computer Vision and Pattern Recognition（CVPR）

论文全称：Hypergraphs for Joint Multi-View Reconstruction and Multi-Object Tracking

论文地址：

Hypergraphs for Joint Multi-view Reconstruction and Multi-object Tracking

论文代码：

https://github.com/neohanju/HYPERGRAPH_TRACKING

地位：

个人理解

• 创新点：只需要一个单一的跟踪图便能解决多摄像头多目标跟踪任务，并且不需要关于跟踪目标的数量的先验知识。
• 为什么：目前已有的多摄像头多目标跟踪算法都是将重建和跟踪作为单独的阶段进行处理，也有将重建和跟踪作为一个全局问题进行优化处理，但都过于繁琐
• 怎么做：建立假设，设计公式来描述多摄像头多目标跟踪问题， 将公式映射到超图，使用超图来解决跟踪问题，对超图添加一些约束条件，将超图转化为约束代价流图，即跟踪和重建问题（多摄像头多目标跟踪任务）被完全定义为一个二进制整数规划问题，最终使用MATLAB bintprog求解器和 faster CPLEX binary integer programming 求解器进行求解。

一、摘要

We generalize the network flflow formulation for multi object tracking to multi-camera setups. In the past, reconstruction of multi-camera data was done as a separate extension. In this work, we present a combined maximum a posteriori (MAP) formulation, which jointly models multi camera reconstruction as well as global temporal data association. A flflow graph is constructed, which tracks objects in 3D world space. The multi-camera reconstruction can be effificiently incorporated as additional constraints on the flflow graph without making the graph unnecessarily large. The fifinal graph is effificiently solved using binary linear programming. On the PETS 2009 dataset we achieve results that signifificantly exceed the current state of the art.

• 本文将多目标跟踪的网络流公式推广到多摄像机设置中，提出了一个组合最大后验(combined maximum a posteriori ,MAP)公式，它将**多摄像机数据的重建（reconstruction）和全局时间数据关联（tracking）**进行了联合建模。