Multiple-target tracking (MTT) is one of the key aspects of driver assistance systems, and has been subject to considerable research. One critical part of an MTT system is solving the data association problem, which is associating correct observations to the existing tracks. Several data association algorithms have been proposed to solve this problem. The multiple hypothesis tracker (MHT) is a well-studied and currently the preferred method of data association in MTT application. MHT maintains several possible data association hypotheses or solutions and uses new observations to eliminate unlikely hypotheses over time. Despite being regarded as the most prominent data association method, MHT implementations remains a challenge because of its computational complexity. In this project, we have described an efficient method for reducing the computational requirement of MHT in automotive applications. This work also presents four most commonly used association algorithms. First, the two types of single hypothesis trackers are introduced: sub-optimal nearest neighbour or SNN and global nearest neighbour or GNN. Then the all-neighbour data association methods: probabilistic data association (PDA) and joint probabilistic data association (JPDA), and their extensions are presented. The algorithms are implemented and tested for four different real case road scenarios, in presence of clutter and missed detection, to check which tracker is more suitable to track which traffic situation. The optimal sub-pattern assignment (OSPA) metric is used to quantify and compare the performance of the association algorithms.
-
Notifications
You must be signed in to change notification settings - Fork 0
License
RahulKumarBazia/Multiple-Target-Tracking
Folders and files
Name | Name | Last commit message | Last commit date | |
---|---|---|---|---|
Repository files navigation
About
No description, website, or topics provided.
Resources
License
Stars
Watchers
Forks
Releases
No releases published
Packages 0
No packages published