CDES: A pixel-based crowd density estimation system for Masjid al-Haram

► CDES is build to estimate and study crowd density at Masjid al-Haram. ► Crowd feature is extracted by using background removal and edge detection method. ► The blob pixels are scaled accordingly to correct perspective distortion. ► Use neural network to estimate the number of people within the blo...

Full description

Saved in:

Bibliographic Details
Published in:	Safety science Vol. 49; no. 6; pp. 824 - 833
Main Authors:	Hussain, Norhaida, Yatim, Halimatul Saadiah Md, Hussain, Nor Liza, Yan, Jasy Liew Suet, Haron, Fazilah
Format:	Journal Article
Language:	English
Published:	Kidlington Elsevier India Pvt Ltd 01-07-2011 Elsevier
Subjects:	Biological and medical sciences BP neural network Crowd density estimation Hajj Masjid al-Haram Medical sciences Miscellaneous People counting Perspective distortion correction Pixel-based feature extraction Public health. Hygiene Public health. Hygiene-occupational medicine Pixel-based feature extraction Crowd density estimation Perspective distortion correction BP neural network Masjid al-Haram People counting Hajj Human Density estimation Image processing Crowd Religion Distortion Neural network Perspective Extraction Counting Corrections Pixel Public health
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	► CDES is build to estimate and study crowd density at Masjid al-Haram. ► Crowd feature is extracted by using background removal and edge detection method. ► The blob pixels are scaled accordingly to correct perspective distortion. ► Use neural network to estimate the number of people within the blob. ► The density is divided into five ranges; very low, low, moderate, high and very high. CDES is an automatic crowd density estimation system that can be used to estimate crowd density from digital images taken at Masjid al-Haram. Developed using a combination of image processing and artificial intelligence (AI) technologies, CDES possesses the capability to count the number of people in moderately high crowds from a flexibly selected region of interest (ROI). Background removal and edge detection are first applied to the image for crowd feature extraction. Then, the extracted crowd foreground blob pixels are scaled accordingly to correct perspective distortion. Finally, the corrected pixel blobs act as input for the backpropagation (BP) neural network to estimate the number of people within the blob. Using the area of the selected ROI, the crowd density is calculated and classified into five ranges from very low to very high. The experimental results are presented.
Bibliography:	ObjectType-Article-1 SourceType-Scholarly Journals-1 ObjectType-Feature-2 content type line 23
ISSN:	0925-7535 1879-1042
DOI:	10.1016/j.ssci.2011.01.005