Vehicle Detection and Tracking for the Gas Station Digital Twin System

Abstract

Abstract Objectives/Scope A digital twin is a virtual representation of a real-world system used to digitally model performance, identify inefficiencies, and design solutions to improve its physical counterpart. With the success of digital twin systems in various scenarios, the digital twin system of gas station is expected to effectively improve the management efficiency and intelligence level of gas stations. Some typical applications include data analysis, traffic prediction, and route planning in gas station, etc. Methods, Procedures, Process The key part of the digital twin system of gas station is how to achieve accurate and efficient vehicle detection and tracking under various environmental conditions. Aiming at this goal, this paper proposes a practical vehicle detection and tracking method for gas station scenarios. We assume the digital twin system consists of a number of digital cameras and precisious 3D models of scene and facilities in gas station. We then present the proposed vehicle detection and tracking pipeline which includes the following three steps: 1) camera internal and external parameters calibration with known scene; 2) vehicle detection and real-time multi-target tracking, and 3) cross-camera re-identification and occlusion handling. Results, Observations, Conclusions We have conducted extensive experiments on several datasets. Experimental results show that the proposed method has good robustness to lighting and occlusion, and can effectively solve the problem of vehicle detection and tracking for the digital twin system of gas station. Novel/Additive Information In a real gas station scenario, real-time tracking of multiple vehicles can be achieved with an accuracy over 90%.