Crowd Density Estimation for Outdoor Environments

Abstract

Crowd density analysis is crucial in management and control of the crowds and ensure safety. In this paper, we proposed two crowd density estimation methods using texture descriptors of the image in outdoor scenes. Two methods based on different techniques, one using Local Binary Pattern and Gabor filters, and the other using 5 statistics of Grey Level Co-occurrence Matrix and firefly classification algorithm, are applied on the dataset. The assessment and comparison with other algorithms have been carried out using different sets of PETS 2009 dataset. The proposed approaches can carry out the estimation more accurately, the rate of true classification is 93.16% and 92.99%, respectively. The results show that our proposed algorithms outperform the other algorithms with a significant margin.