Modeling from an Object and Multi-object Tracking System

Abstract

This paper describes an implementation system tracking and detection object and multi-object. We trait the different tracking algorithms of moving objects in video sequences. The tracking a particular precise object, robust, efficient, reliable and fast is a difficult problem to solve. Hence, it is a very essential for computer vision applications. For this reason, tracking objects in video sequences is a very active area of research since the 1970. It has attracted many people in the area of research for use in a variety of applications. The tracking is often the first step in an analysis of the activities, detection, behavior, interactions and relationships between objects of interest. Many methods of tracking objects have been proposed and developed. Tracking is the estimation and analysis of the trajectories of an object in the frame plane by moving in an frame where video sequence. Most of the motion objects tracking algorithms take input frames captured by a fixed camera to give at output a video sequence. However, there are other algorithms that take as input a video sequence to provide a video at output containing the tracked objects. These algorithms perform a first step of the detection objects in order to determine which of the pixels of the current frame which belong to the background of the sequence and which represent the motion objects. The problem of object tracking can be expressed in terms of detection of the object in each frame of the video sequence. The set of applications (security systems, military systems, intelligent transportation systems, video conferencing, surveillance, etc.) can be addressed. So, the lightness of the systems also lets you open recent applications such as human-machine interface and/or the human-robot interface, where the sensors can be embedded on a mobile robot. This usually requires the exploitation of object tracking techniques and algorithms in motion that will register and implement on FPGA targets that are introduced in these interfaces. We can cite as an example then the main existing algorithms in the literature as the block-matching, the KLT, the Kalman filter, the Meanshift and the Camshift.