Pistage d'objets multiples dans le cas d'un lidar à faible résolution angulaire

Authors: Roy-Labbé, Maude
Advisor: Grenier, Dominic; Lamontagne, Luc
Abstract: This thesis presents an analysis of the performance of tracking algorithms in the case of a lidar with low angular resolution. More particularly, we are interested in a lidar system composed of individual sensors each covering a distinct angular region. The sensors used have the particularity of measuring only the distance of the objects encountered, thus limiting the angular resolution to their beam. The algorithms were tested using simulation data based on the lidar system. In the case of simple detections, an instant tracking algorithm based on a Kalman filter was more than sufficient. For more complex cases, the use of multiple hypothesis theory (MHT) made it possible to improve tracking results. In this method, when there is an ambiguity in the tracking, the possible hypotheses are considered simultaneously until the information received at subsequent times makes it possible to identify the correct one. For the case under study, optimal results were obtained for an MHT considering at most 3 hypotheses at any time and waiting at most 3 time steps to make a decision. Overall, the algorithms presented reacted better to false alarms rather than to non-detections. A method for optimizing the calculation times of the algorithms has also been developed. This method is based on Murty's algorithm goes from a simple and fast tracking method (here by nearest neighbors) to a more complex association method (here using a Kalman lter) only when an ambiguity is detected. In the case of a simple situation with two targets, performances comparable to that of an association by Kalman filter were obtained with a calculation time of less than 10% of that usually required.
Document Type: Mémoire de maîtrise
Issue Date: 2020
Open Access Date: 7 June 2021
7 June 2021
Permalink: http://hdl.handle.net/20.500.11794/69303
Grantor: Université Laval
Collection:Thèses et mémoires

Files in this item:
Description SizeFormat 
36303.pdf1.67 MBAdobe PDFThumbnail
36303_Annexe.zip573.2 kBArchive ZIPView/Open
36303_Annexe_Contenu.txt967 BTextView/Open
All documents in CorpusUL are protected by Copyright Act of Canada.