MAS for video objects segmentation and tracking based on active contours and SURF descriptor

TL;DR

This paper introduces a novel multi-agent system-based algorithm utilizing SURF features for robust and efficient video object segmentation and tracking, inspired by active contours, implemented on a parallel computing platform.

Contribution

It proposes a new multi-agent system approach combining active contours and SURF descriptors for improved video segmentation and tracking.

Findings

Algorithm is more robust than previous methods.

It achieves faster processing speeds.

Experimental results validate effectiveness.

Abstract

In computer vision, video segmentation and tracking is an important challenging issue. In this paper, we describe a new video sequences segmentation and tracking algorithm based on MAS "multi-agent systems" and SURF "Speeded Up Robust Features". Our approach consists in modelling a multi-agent system for segmenting the first image from a video sequence and tracking objects in the video sequences. The used agents are supervisor and explorator agents, they are communicating between them and they inspire in their behavior from active contours approaches. The tracking of objects is based on SURF descriptors "Speed Up Robust Features". We used the DIMA platform and "API Ateji PX" (an extension of the Java language to facilitate parallel programming on heterogeneous architectures) to implement this algorithm. The experimental results indicate that the proposed algorithm is more robust and faster than previous approaches.