Towards an Evaluation and Repair Framework for Video Interpretation

LES AUTEURS Benoit Georis

RESUME:

In the past few years, many interpretation systems have been developed but none of them have been successfully applied to real world applications. One major weakness of these systems is the
tracking process. Tracking is still a central issue in scene interpretation, as the loss of a tracked object prevents the analysis of its behaviour. Tracking has been extensively studied for many years. Various techniques have been explored, both model-based and model-free. Nevertheless, the tracking problem remains unsolved since there are many sources of ambiguities like shadows, illumination changes, over-segmentation and mis-detection. These difficulties need to be handled in order to make the correct matching decision.
In addition, the increasing number of installed surveillance systems need reasoning capabilities requiring highly efficient tracking algorithms. These systems are running 24 hours a day in varying conditions. Our goal is to conceive a generic human tracking algorithm which can adapt itself automatically to a scene change.
To face this problem, we propose in this work a general framework for algorithm evaluation, applied to the tracking problem. As a first step, we investigate supervised evaluation since we compare
algorithm outputs with ground truth. Our final objective is to have a fully automatic evaluation to adapt dynamically the interpretation platform to any new situations. Algorithm assessment is mandatory - but not sufficient - to design robust systems. We first propose a global evaluation method for ranking tracking algorithms. Evaluation criteria have been defined accordingly. Second, we propose a fine evaluation method which classifies and diagnoses tracking errors. This step is algorithm dependant and try to locate exactly the code or the parameter which is responsible for an error class. This second evaluation level and the associate repair methodology is the main contribution of this work.

Mots clé: human tracking, performance assessment, supervised evaluation, program
supervision

Pour télécharger cet article,  cliquez ici

BibTeX reference:

BibTex:

@MISC{Georis03,
AUTHOR = {B. Georis},
TITLE = {Towards an Evaluation and Repair Framework for Video
Interpretation},
NOTE = {Master Thesis, Université Catholique de Louvain},
YEAR = 2003
}