Géolocalisation des objets détectés par une caméra monoculaire à bord d’un véhicule

Abstract

This work is being carried out as part of a research project funded by the Quebec Ministry of Transport on data acquisition for autonomous vehicles. Specifically, the aspect of geolocating objects surrounding the vehicle will be dealt with. So the aim of this project is to address the challenge of geolocating objects detected by a monocular camera mounted on a vehicle, with the goal of automating the GPS-based localization of road anomalies such as faulty streetlights and potholes. The main objective is to map Quebec by identifying the GPS positions of objects of interest, in order to create a database that could be used for various applications, such as automating the process of identifying road anomalies or providing information to autonomous vehicles about the location of anomalies such as potholes or traffic lights. The localization methods based on orientation lines were chosen due to their simplicity and their low demand in terms of computation time. Detection, tracking, and geolocation algorithms were developed to generate orientation lines and estimate the GPS positions of detected objects. Experimental results showed 88% accuracy within a 20-meter radius for geolocating streetlights. Improvements are possible to strengthen the algorithm’s performance, particularly by calibrating the camera’s pose and improving the GPS configuration.

Date	27 Apr 2023
Original language	French
Awarding Institution	École de technologie supérieure
Supervisor	Maarouf Saad (Supervisor) & Pier Marc Comtois-Rivet (Co-supervisor)