Abstract Multi-view video surveillance is a highly valuable tool to ensure the safety of the crowd in large public space. By utilizing complementary information captured by multiple cameras, the issue of limited views and occlusion in single views can be addressed to gain better insight into the whole monitored space. However, multi-view surveillance has been widely applied to microscopic crowd analysis, for example pedestrian detection and tracking, while macroscopic level analysis, which deals with the whole crowd, has received little attention. We propose a multi-view framework for the generation of level of service maps, which are the most commonly used measure of congestion at macroscopic level, based on an ensemble of state-of-the-art Convolutional Neural Networks (CNNs). Several combination rules are compared and evaluated on two datasets, both in sparse and dense scenarios. Our results show that this fusion framework improves the accuracy of level of service map generation, from 83.2% to 89.8%, and eliminates blind spots in single views. Our framework is implemented on a 3 D GIS platform, which provides a suitable interface for multi-view crowd congestion management. The results of a loading test show that a maximum of 48 cameras can be processed at a map refresh rate of 2 seconds.

中文翻译：

---

基于卷积神经网络分类器集成的多视图人群拥堵监控系统

摘要 多视角视频监控是保障大型公共空间人群安全的一种极具价值的工具。通过利用多个摄像头捕获的互补信息，可以解决单个视图中的有限视图和遮挡问题，从而更好地了解整个监控空间。然而，多视图监控已广泛应用于微观人群分析，例如行人检测和跟踪，而针对整个人群的宏观层面分析却很少受到关注。我们提出了一种用于生成服务地图级别的多视图框架，这是基于最先进的卷积神经网络 (CNN) 的集合，这是宏观级别最常用的拥塞度量。在两个数据集上比较和评估几个组合规则，在稀疏和密集场景中。我们的结果表明，该融合框架将服务地图生成水平的准确性从 83.2% 提高到 89.8%，并消除了单视图中的盲点。我们的框架是在 3D GIS 平台上实现的，该平台为多视图人群拥塞管理提供了合适的接口。加载测试结果表明，在2秒的地图刷新率下，最多可以处理48个摄像头。