Roadside Vehicle People Counter

Offentlig prosjektsammendrag

In this project we are developing a camera-based sensor system capable of counting and classifying people in passing vehicles. Project is targeted for vehicle ferries applications, with tolling, carpooling, parking and security as secondary opportunity markets. A successfully developed solution will improve the safety precautions on Norwegian vehicle ferries. Currently the passengers in vehicles are manually counted when cars are driving onboard. This returns inaccurate results. The proposed project is based on performed field tests of existing technology and solutions where we have experienced that remote video counting through car windows in changing light conditions is a difficult task which needs further research and development. Different vehicles types, changing light, changing weather, tinted windows and coloured glass makes it difficult to see how many people are inside each vehicle. In this project we will address this imaging challenge and also develop computer vision algorithms to detect people from several different camera viewpoints, using innovative camera solutions and data processing based on deep learning machine vision technology.