Counting passengers and vehicles with computer vision techniques for city of Edmonton traffic

轻轨交通(LRT)的乘客数量和道路交叉口的车辆分类是埃德蒙顿市用于交通和交通规划的重要交通数据。目前，数据采集是人工进行的，而且一年只进行一次，因为人工采集数据的成本很高。在这项提议中，我们提出了一种基于计算机视觉的低成本解决方案，用于这些数据收集，这种解决方案将以非惯性和自动化的方式全天候进行。我们的解决方案将使用安装在轻轨车站和道路交叉口的现有摄像机，将总成本降低到仅限于算法开发及其配备图形处理单元(GPU)的个人计算机的部署。在计算机视觉方面，我们提出的解决方案将使用最先进的算法及其在乘客计数和车辆分类问题上的进一步发展。

Passenger count for Light Rail Transit (LRT) and vehicle classification at the road intersections are important transit data that the City of Edmonton uses for transit and transportation planning. At present, the data collection is manual and it occurs only once in a year, because the cost of manual data collection is high. In this proposal we present a computer vision-based low-cost solution to these data collections that will occur 24/7 in a non-intrisive and automated manner. Our solution will use existing video cameras mounted at LRT stations and road intersection bringing the overall cost down only to the algorithm development and its deployment with personal computers equipped with graphics processing units (GPUs). On the computer vision side, our proposed solution will use promising state-of-the-art algorithms and their further development to the passenger counting and vehicle classification problems.