Computer vision-based condition assessment of the public transit infrastructure assets

点击翻译按钮获取中文摘要

Public transit systems are essential to the social and economical prosperity of Canada, and contribute to the advancement of social equity and environmental objectives. Efficient and effective infrastructure asset management is vital to providing a reliable public transit system, but current infrastructure asset management practices mainly rely on conventional data gathering methods which require manual site visits that are labour-intensive and time-consuming. Effective asset management could be more challenging for small and medium-sized cities due to unsteady and limited public revenues. Since transit agencies have difficulties in maintaining and updating their large datasets using their limited human resources, many research efforts have focused on development of semi-automated infrastructure condition assessment methods, such as crowdsourcing by the online community, image and video processing, aerial photography, and vehicle-based LiDAR. The proposed research project will develop an innovative computer vision-based system to improve data collection and condition assessment processes for public transit infrastructure asset management. The proposed system will use the regular operation of the transit buses for data collection and then utilizes deep learning methods to detect public transit infrastructure assets, asses their condition, and update asset management inventories. The industry partners in this research project are Thunder Bay Transit and Consat Telematics. Thunder Bay Transit provides general public transit services in Thunder Bay area and aims to incorporate the findings of this research to enhance management of their infrastructure assets. Consat Telematics is a technology company which develops intelligent transportation systems and the developed system will be integrated into their fleet management systems.