Human Tracking Using Weighted Running Window Background Model Based GMM

Volume 12 Issue 2 October - December 2017

Research Paper

Human Tracking Using Weighted Running Window Background Model Based GMM

Harihara Santosh Dadi*, Gopala Krishna Mohan Pillutla**, Madhavilatha Makkena***

* PhD Scholar, Department of Electronics and Communication Engineering, JNT University, Hyderabad, India.

** Professor, Institute of Aeronautical College of Engineering, Hyderabad, India.

*** Professor, Department of Electronics and Communication Engineering, JNT University, Hyderabad, India.

Dadi., Pillutla., and Makkena. (2017). Human Tracking Using Weighted Running Window Background Model Based GMM. i-manager’s Journal on Software Engineering, 12(2), 44-54. https://doi.org/10.26634/jse.12.2.14066

Abstract

Tracking of humans in video streams is important for many applications. For tracking purposes, many algorithms have come up in the recent years. The most prominent one among all of them is Gaussian Mixture Model (GMM). This algorithm is basically employed for tracking the objects in the Video scene. Later, the algorithm has been modified for the purpose of tracking humans. GMM uses only single rectangular template for tracking an object. In order to track humans specifically, the template has been divided into four regions. The top region is for the head and the remaining regions are for the chest, waist and legs respectively. All the regions are of rectangle shape. Connection has been established among all the regions assuming that all four regions will move at a time for humans. There is only 10% horizontal variation allowed between the regions. The proposed algorithm could handle both partial occlusion and full occlusion. The new algorithm is compared with the tracking system of GMM algorithm. The precision, recall, false alarm per frame, false negatives, false positives and mostly lost are compared with the existing GMM. The time taken for processing a single frame is reduced by using new algorithm when compared with the existing algorithm. Performance metrics show that the Weighted Running Window Background (WRWB) Model Based GMM algorithm out performs when compared with GMM algorithm in terms of time taking.