Computer Science > Computer Vision and Pattern Recognition
[Submitted on 8 Nov 2017]
Title:Heuristic Search for Structural Constraints in Data Association
View PDFAbstract:The research on multi-object tracking (MOT) is essentially to solve for the data association assignment, the core of which is to design the association cost as discriminative as possible. Generally speaking, the match ambiguities caused by similar appearances of objects and the moving cameras make the data association perplexing and challenging. In this paper, we propose a new heuristic method to search for structural constraints (HSSC) of multiple targets when solving the problem of online multi-object tracking. We believe that the internal structure among multiple targets in the adjacent frames could remain constant and stable even though the video sequences are captured by a moving camera. As a result, the structural constraints are able to cut down the match ambiguities caused by the moving cameras as well as similar appearances of the tracked objects. The proposed heuristic method aims to obtain a maximum match set under the minimum structural cost for each available match pair, which can be integrated with the raw association costs and make them more elaborate and discriminative compared with other approaches. In addition, this paper presents a new method to recover missing targets by minimizing the cost function generated from both motion and structure cues. Our online multi-object tracking (MOT) algorithm based on HSSC has achieved the multi-object tracking accuracy (MOTA) of 25.0 on the public dataset 2DMOT2015[1].
References & Citations
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Connected Papers (What is Connected Papers?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
alphaXiv (What is alphaXiv?)
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Hugging Face (What is Huggingface?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
CORE Recommender (What is CORE?)
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.