Multitarget Tracking in Nonoverlapping Cameras Using a Reference Set

Abstract

Tracking multiple targets in nonoverlapping cameras are challenging since the observations of the same targets are often separated by time and space. There might be significant appearance change of a target across camera views caused by variations in illumination conditions, poses, and camera imaging characteristics. Consequently, the same target may appear very different in two cameras. Therefore, associating tracks in different camera views directly based on their appearance similarity is difficult and prone to error. In most previous methods, the appearance similarity is computed either using color histograms or based on pretrained brightness transfer function that maps color between cameras. In this paper, a novel reference set based appearance model is proposed to improve multitarget tracking in a network of nonoverlapping cameras. Contrary to previous work, a reference set is constructed for a pair of cameras, containing subjects appearing in both camera views. For track association, instead of directly comparing the appearance of two targets in different camera views, they are compared indirectly via the reference set. Besides global color histograms, texture and shape features are extracted at different locations of a target, and AdaBoost is used to learn the discriminative power of each feature. The effectiveness of the proposed method over the state of the art on two challenging real-world multicamera video data sets is demonstrated by thorough experiments.

Multi-target tracking using predefined reference set in Non-Overlapping Cameras by AdaBoost algorithm

“…Due to variation in illumination conditions, and camera imaging characteristics, there might be significant appearance change of target across camera views. In multitarget tracking system there is need to track same object across multiple camera output tracks provided cameras are consist of non-overlapping regions [2]. Consequently the same target may appear in very different in two cameras which causes to error.…”

“…This project proposes reference set based appearance of the target to determine the similarity of multiple targets in different camera. The goal of multi-tracking across camera is to associate tracks in different cameras which contain the same object [2].Our main goal is to implement the algorithm for multi-target tracking in non-overlapping cameras using predefined reference set.…”

“…The contributions of this paper are as follows  The paper proposes the work for providing solution to track same target across multi-cam dataset [2].  A novel reference set based appearance model to improve multi-target tracking in a network of non-overlapping camera.…”

“…Nowadays CCTV cameras are used for many purposes such as in public places, schools, airports, colleges etc. For security purpose the rate of use of CCTV cameras are increases [2]. The paper proposes to track the object robustly using multi-target tracking concept in non-overlapping field of views.…”

“…Some previous methods are used for multi-target tracking that are color histogram, brightness transfer function (BTF) [11]. Many times it is not possible to cover complete area of interest by using single camera, such a cases there is need to use multi-target tracking system with non-overlapping field of views (FOV) [2]. In this paper we use method of feature extractions that is AdaBoost.…”