1D Signal Processing for Improvement of People Counting Estimation Results

The paper proposes a novel method to overcome the problem of occlusion while estimating the density of pilgrims in highly congested scenarios. Millions of Muslims gather at Al-Haram Mosque every year to perform Hajj and Umrah. Therefore, the need to develop a robust computer aided system for the estimation of density of pilgrims for averting stampedes can not be denied. The proposed method uses the number of motion vectors with non-zero magnitude for designing a linear regression model to estimate the density of pilgrims. There are several challenges associated with the development of such a system. Therefore the number of motion vectors do not always linearly correlate with the corresponding number of pilgrims. The property of inter-frame correlation is exploited to maintain linearity between the number of motion vectors and people count using smoothing filter from 1D signal processing techniques. Experimental results show that the Percentage Mean Absolute Relative Error (PMARE) of the proposed system is nearly 2.5%.