Crowd Density Estimation Based on ELM Learning Algorithm

Abstract

Crowd density estimation in public areas with people gathering and waiting is the important content of intelligent crowd surveillance. A real-time and high accuracy algorithm is necessary to be inputted in the classification and regression of crowd density estimation to improve the speed and increase the efficiency. Extreme Learning Machine (ELM) is a neural network architecture in which hidden layer weights are randomly chosen and output layer weights determined analytically. In this paper, we propose a new method which is based on Haralick's texture vectors, Gray-level Co-occurrence Matrix and ELM. The datasets are based on PETS2009 and UCSD. The performances are compared among SVM, BP and ELM. The experimental results suggest that the ELM learning algorithm has a good performance of accuracy and a very fast speed than other methods.

Real-Time, Embedded Scene Invariant Crowd Counting Using Scale-Normalized Histogram of Moving Gradients (HoMG)

“…This ever-growing interest in automated crowd analysis has prompted researchers to propose a variety of methods for video-driven crowd segmentation and counting [16,17,20,22,29,4,6,13,28], with a recent survey on state-ofthe-art methods found in [25]. However, widespread industrial adoption of automated crowd segmentation and counting requires a simple, low-cost, highly-scalable deployment process, which is difficult to achieve using existing methods.…”