A Probabilistic Approach for Detecting Human Motion in Video Sequence using Gaussian Mixture Model
Nazia Aslam, Maheshkumar H. Kolekar · 2022 2nd International Conference on Emerging Frontiers in Electrical and Electronic Technologies (ICEFEET) · 2022
The detection and recognition of the human body in motion are two essential and demanding problems in visual surveillance systems. The detection process is slow and chal-lenging due to the dynamic nature of the background and the occlusion of moving objects. This paper presents a probabilistic approach to model the dynamic behaviour of the background and accordingly detect multiple moving human bodies in a video sequence. The Gaussian Mixture Model (GMM) is used for adaptive background modelling. Blobs are drawn on the most significant and moving clusters of pixels in a binary image obtained after background subtraction to detect and track the moving humans. This simplified approach results in forming a pattern of various structural blobs. Finally, we demonstrate the effectiveness of our proposed approach with many tests on the AVG- TownCentre and PETS09-S2L1 datasets. The detection speed of the proposed algorithm is 30 fps and 32 fps for AVG-TownCentre and PETS09-S2L1 datasets, respectively, have been recorded. Also, we have reported a less computation time of 3.8ms for the proposed model.