[1]
presents the use of crowd simulation for emergency response using BDI agents based on immersive virtual reality. The results indicate that constructed simulation can be used as an effective emergency management tool. Reference [2] presents prediction system of human crowd pressures. Utilising the standard forward-backward autoregressive modelling techniques for spectral analysis of a measured signal, predictions of pressures generated by very high densities of pedestrians have been formulated. The study suggests that several minutes are available for corrective action to be taken to avoid an accident involving crushing of pedestrians. In [3], crowd monitoring using image processing is presented. The paper presents some image processing techniques of visual images for data collection and on-line monitoring of crowds. In general, more research is still needed in the area of on-line monitoring to estimate people's density and prevent accidents. The demand for intelligent and integrated monitoring systems of people has significant important during Hajj. Hajj (the annual Muslim pilgrimage to Makkah (Mecca), Saudi Arabia) is a large-scale event that currently attracts more than two millions of visitors every year from all over the world for about two weeks. It is described as the biggest annual mass movement of people on the planet [4], and it is expected to attract 4. 8 millions people by year 2025. The paper describes the use of infrared, light and temperature sensors to investigate the design of a crowd monitoring system for Madina Mosque in Saudi Arabia which most people visit before or after the Hajj activity. System Overview The development of an autonomous monitoring system technology is an essential enabling technology that can provide the means for capturing and integrating information. Fig. 1-a presents the autonomous monitoring system that is developed for crowd monitoring. The system includes a monochrome visual camera attached to a small computer system via a video capture card type National Instruments NI PCI-1407. A low cost infrared camera type IRISYS IRI1002 is connected to the same computer system using an RS232 connection. Temperature and light sensors, designed by IENSYS Ltd [5], are interfaced to the computer system via IENSYS board [6] to the RS232 port of the system. The computer system is Ethernet enabled so that data can be transferred to central analysis point as described in Fig. 1-a. RS232 Infrared System (IRI1002) Monochrome Camera Temperature Sensor Light Intensity Sensor Video Capture Card Computer system Data Capturing Software Data Capturing Software NI PCI-1407 NI PCI-1407 Internet IensysBoard.
Google Scholar
[27]
[28] [29] [30] [31] [32] [33] [34] [27] [28] [29] [30] [31] [32] [33] [34] Infrared image Visual image.
Google Scholar
[27]
[28] [29] [30] [31] [32] [33] [34] [27] [28] [29] [30] [31] [32] [33] [34] Infrared image Visual image.
Google Scholar
[27]
[28] [29] [30] [31] [32] [33] [34] [27] [28] [29] [30] [31] [32] [33] [34] Infrared image Visual image (b) 27.
Google Scholar
[28]
[29] [30] [31] [32] [33] [34] [27] [28] [29] [30] [31] [32] [33] [34] Infrared image Visual image.
Google Scholar
[27]
[28] [29] [30] [31] [32] [33] [34] [27] [28] [29] [30] [31] [32] [33] [34] Infrared image Visual image.
Google Scholar
[27]
[28] [29] [30] [31] [32] [33] [34] [27] [28] [29] [30] [31] [32] [33] [34] Infrared image Visual image.
Google Scholar
[27]
[28] [29] [30] [31] [32] [33] [34] [27] [28] [29] [30] [31] [32] [33] [34] Infrared image Visual image (b) 27. 5.
Google Scholar
[32]
32. 5 27. 5.
Google Scholar
[32]
32. 5 Infrared image Visual image 27. 5.
Google Scholar
[32]
32. 5 27. 5.
Google Scholar
[32]
32. 5 Infrared image Visual image 27. 5.
Google Scholar
[32]
32. 5 27. 5.
Google Scholar
[32]
32. 5 Infrared image Visual image (c) (a) Fig. 1: The autonomous data capturing system (a) and examples of the images captured (b and c). The Experimental Work The Experimental work of data capturing described in this paper has been performed at the Madina Mosque in Saudi Arabia. The system is installed to view people from a high corner as shown in Fig. 1-b and Fig. 1-c. Data is captured for different densities of people during a period of over one month. Every 1 minute a frame of data is captured and saved to the computer for analysis. The selected view is realistic and has many sources of noise such as the sunlight and air conditioning systems. Fig. 1-b presents the infrared image and the associated visual image at the Madina Mosque during praying. Notice that the infrared image is warm, in a specific linear pattern based on the lines of people praying. After prying people start moving randomly and this is also being shown in the infrared image, Fig. 1-c. Fig. 1-b and Fig. 1-c present the raw data that the system need to analyse in order to be able to predict the density of people. One of the common problems is that people closer to the camera will look warmer and larger in size (i. e. will occupy more number of pixels). People in the far-side of the mosque will look smaller and at less temperature level. A calibration process will be needed to normalised the data based on the geometry of the view. Results and Discussion Raw Data Captured. Fig. 2 presents an example of the captured data for a full day. The first curve represents the average infrared data of the infrared image that is captured every 1 minute period. The second graph represents the number of pixels above a specific threshold which represents human temperature. The average value is about 30 o C, but it is mainly dependent on the distance, ambient temperature and the geometry of the view. The third curve represents the intensity of light measured by a light sensor. Notice that at image (a) the mosque is empty before sun rise and this is indicated in low level of infrared and light intensity. At image (b), the sun starts to rise, notice that the light sensor is detecting increase in the light values. This is also being reflected in gradual increase in the infrared radiation. In this paper, the temperature was constant due to the air conditioning system inside the building.
Google Scholar
[20]
[30] [40] Average infrared 500 1000 Light.
Google Scholar
500 1000 1500.
Google Scholar
200 400 Infrared Pixels Frame number (a) (b) (c) (d).
Google Scholar
[20]
[30] [40] Average infrared 500 1000 Light.
Google Scholar
500 1000 1500.
Google Scholar
500 1000 1500.
Google Scholar
200 400 Infrared Pixels Frame number (a) (b) (c) (d) Fig. 2: The density of people as predicted by different sensors (with no signal processing). Notice the four peaks in the average infrared radiation and number of warm infrared pixels that represents Zuhr, Assr, Magrib and Isha Prayers. The infrared curves indicate that Magrib and Isha prayers have maximum densities. However, it is evident that Isha prayer takes longer time. By focusing on the warm infrared pixels, Fig. 3 presents the expected densities and the associated visual images. The number of warm infrared pixels are found to be very representative to the density of people in the mosque. However, there is a need to compensate in that location for the reflection of light during the early hours of the morning, image (b).
DOI: 10.1109/icicip.2013.6568029
Google Scholar
500 1000 1500.
Google Scholar
[50] 100 150 200 250 Frame number.
Google Scholar
500 1000 1500.
Google Scholar
[50] 100 150 200 250 Frame number Infrared Pixels (a) (b) (c) (d) (e) (f) (g) Fig. 3: The number of warm infrared pixels without compensation. Sensor Fusion Model. Sensor fusion model is developed to compensate for the effect of light intensity on the infrared image. Equation 1 presents the methodology used to compensate for light variation. xλρφ −= (1) Where φ is the compensated number of infrared pixels above a specific threshold value, ρ is the initial number of infrared pixels, x is the light intensity, and λis a calibration constant. Fig. 4 presents the number of infrared pixels (φ) after compensating for the intensity of light. Notice that that noise in the signal has been removed at image (b) of Fig. 4. This has caused the number of infrared pixels (i. e. warm pixels) to be very accurate representation to the density of people. The temperature sensor has not been used in this work since the Mosque has an air conditioning system which maintains the building at a narrow range of temperatures. Fig. 4 shows very clear indication of the density and the occupancy of the mosque over 24 hours period. The system can be utilised for monitoring the density and behaviour of people in different locations in the mosque to provide comprehensive on-line crowd monitoring system.
Google Scholar
500 1000 1500 -100.
Google Scholar
100 200 300 Frame number Infrared Pixels (compensated).
Google Scholar
500 1000 1500 -100.
Google Scholar
100 200 300 (compensated) (a) (b) (c) (d) (e) (f) (g) Fig. 4: By compensating for the light intensity, the infrared data has shown accurate results. Conclusion The novel application of low-cost infrared system has been found to be very successful in estimating people's density in Madina Mosque. The change in light intensity has caused significant source of error. Sensor fusion of infrared data and light intensity data has been found useful in compensating for the error and proving accurate results. Future work will investigate the use of visual image processing in comparison to the infrared data. References.
DOI: 10.7554/elife.28342.003
Google Scholar
[1]
] Ameya Shendarkar, Karthik Vasudevan, Seungho Lee, Young-Jun Son: Simulation Modelling Practice and Theory, Volume 16, Issue 9, October 2008, pp.1415-1429.
DOI: 10.1016/j.simpat.2008.07.004
Google Scholar
[2]
Ris S.C. Lee, Roger L. Hughes: Accident Analysis & Prevention, Volume 38, Issue 4, July 2006, pp.712-722.
Google Scholar
[3]
Anthony C. Davies, Jia Hong Yin and Sergio A. Velastin: IEE Electronic and Communications Engineering Journal, Vol. 7, No. 1 (Feb), pp.37-47.
Google Scholar
[4]
BBC website, accessed May 2009, http: /news. bbc. co. uk/1/hi/world/middle_east/1830877. stm.
Google Scholar
[5]
Iensys Ltd web site, http: /www. iensys. com, accessed on 1 April (2009).
Google Scholar
[6]
A. Al-Habaibeh, R. Cai, M.R. Jackson and R.M. Parkin: Proc. of 7th Int. Conference on Monitoring And Automatic Supervision In Manufacturing, Zakopane, Poland, 19 - 21 August 2004, p.49.
Google Scholar
