Vision Based Railway Platform Monitoring System Literature Review

3870 words (15 pages) Essay in Transportation

23/09/19 Transportation Reference this

Disclaimer: This work has been submitted by a student. This is not an example of the work produced by our Essay Writing Service. You can view samples of our professional work here.

Any opinions, findings, conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of UK Essays.

 

Table of Contents

ABSTRACT

1. INTRODUCTION

2. TECHNICAL OVERVIEW

3. TRAIN DETECTION

4. OBJECT DETECTION

5. FALSE ALARM FLITERING

6. NON CRITICAL ALARM

7. CRITICAL ALARM

8. ADVANTAGES / MAIN LIMITATION

9. CONCLUSION

References

VISION BASED RAILWAY PLATFORM MONITORING SYSTEM

 

ABSTRACT

Railway platform safety in Melbourne has become a prime concern and needs an immediate attention as dozens of people are killed every year by falling from train platform. Vision based platform monitoring system is proposed for the safety improvements. Main purpose of this system is to perceive dangerous situation in the platform and railway tracks such as fallen passenger or dangerous object and to inform the central control, station employee and train driver about the situation by the use of image processing technology. The detection algorithm of the proposed system exploits stereo vision algorithm to improve system detection performance because railway platform has various illumination condition due to train arrival and departure in the scene. The detection process analyses scene and detects both four different train status and fallen objects in present monitoring area. To verify this system extensive experiment were conducted in a metro station. This system will play a key role for establishing intelligent monitoring system in railway platform

1.    INTRODUCTION

For past few centuries, railways have become easiest and popular way of public transportation. With about millions of people boarding the train daily, safety has always been the primary concern. There has been number of incident reported of people falling off the platform to the railway track either by accident, criminal activity or even a suicide attempt. Though the exact number of death toll is unknown but it said that about every eight days someone dies on Victorian rail network [1]. Screen doors or detection mats may be used in metro stations to prevent falling accidents; however, the cost of installation is very high, and the screen door is not appropriate for outdoor train stations. A number of vision based surveillance systems have been proposed to confirm safe railway operation [2-9]. These systems are focused on measuring the level of crowding on the passenger platforms of train services. If overcrowding is detected, the system will notify the station operators to take appropriate actions to prevent dangerous situations such as people falling off or being pushed onto the tracks [10-13]. Background subtraction is commonly used to locate passengers and thereby to estimate the crowd density [2-5, 8] Edge information may be used to measure the regions occupied by passengers [2, 5]. In addition, motion information may be utilized to detect and track moving objects in Ref. [6,7,9]. To detect abrupt and dangerous situations, object tracking has been implemented using a combined method of background subtraction, edge detection and motion detection [10-12]. An infrared camera was found to be useful to locate passengers [11].

The visual surveillance system based on crowd estimation may be useful for detecting potentially dangerous situations. The system, however, may yield too many false alarms because it is not designed to determine critically dangerous situations such as when a passenger falls from the station platform, when a passenger walks on the rail tracks and when a passenger is trapped by a door of a moving train. Especially passengers who have fallen from the station platform cause serious safety problems.

2.    TECHNICAL OVERVIEW

 

VBPMS is an intelligent monitoring system which detects the dangerous situation automatically without constant monitoring of human [14]. This system is useful not only to detect human and object on the railway track but also the passengers stuck between the doors, passengers beyond the safety line, over crowd of people in certain area, passenger stuck between the train and platform gap and so on [14]. This type of system uses video camera, thermal cameras, infrared sensors or even the combination of these sensors, on different location of the railway platform where each component has its own specific area to cover [14-15]. Each of these components checks for any dangerous situation in real time scenario and sends the result accordingly. If any hazardous situation has occurred, then the video of the particular hazard along with waring signal is sent to traffic control, station emergency employee and train driver so that any unpleasant situation could be avoided [14-17].

Fig.1. Concept of Vision Based Platform Monitoring System. [14]

Commonly most of the research on vision based platform monitoring system is divided into three different units; Information acquisition unit, central data fusion unit and information multicasting unit. In information acquisition unit any dangerous situation is detected with the help of thermal cameras and infrared sensors, which are highly reliable in case of object detection. In central data fusion unit all the data from the camera and sensors are collected and analysed. Further to this process, different kind of warning signals and messages are generated by this unit. Finally, information multicasting unit provides different department with the resultant warning message along with incident video and standard operation procedure [10-16].

Fig 2: System Configuration of VBPMS. [14]

For the detection of the object or human, main thing to be considered it that the train itself is not to be taken as dangerous object in the track, if this case is not considered then there will be false alarm every time any train approaches the platform. Another thing to taken in account will the size of the object. If the size of the object is too small, it could be counted as non-dangerous object. These can be done by using various image processing techniques such as thresholding, frame difference, background subtraction, edge detection and so on [17-19]. Flow chart of detection process is as shown below.

Fig 3: Flow chart of detection process. [14]

There will be various warning and alarm associated with this system. This system is capable of generating both critical and non-critical alarm system. For example, if a person is too close to the edge of the platform then the system will send a non-critical waring but if the person falls off the platform from that position then the cameras and sensors are able to detect the fall and will generate a critical warning signal [10].

This system is not only used in the normal railway station but can also be used in underground stations and is very useful in case of automated trains where there are no train drivers [18]. This system will help the automated trains for obstacle detection and ensure its safety. In present scenario where CCTV cameras are used in most of the stations, it is hard to ensure the safety of the passengers as CCTV cameras have limited capacity and human need to constantly monitor it, in case of emergency this might not be efficient therefore VBPMS is comparatively far better technology to ensure the safety of the passengers.

3.    TRAIN DETECTION

Both trains and fallen persons are observed on the tracks, train position estimation is quite important for safety systems. While analysing hazardous situation, determination of train status is also important because train movements must not be regarded as persons’ movements [17]. To make decision of dangerous factor for fallen object in monitoring area, it is important to find the accurate train states in the area for every single stereo camera. Train and monitoring areas should be clearly defined. Generally, as shown in Table 1, a train at the station has four different sates, i.e. approaching state(IN-state), stopping state(ON-state), pulling out state(OUT-state), empty state(OFF-state) [15].

Train States

Description

OFF

There is no train.

IN

Train is approaching.

ON

Train is arrived and stopped.

OUT

Train is pulling out

Table 1. Train states at the station [15]

4.    OBJECT DETECTION

Various method has been purposed for the obstacle detection, one being in which the system calculates distance between object and observing camera, and object height by using image processing technology. To find out distance information the system adopts stereo vision technique. The basic concept of stereo vision is exploiting the similarities along the disparity, the displacement between two separated views. The system searches the minimum displacement in target view for each block of reference view. With the minimum displacement, we can find out distance from camera focus to the target object [20].

The matching block with minimum sum of absolute displacement (SAD) is defined as:

MB(x,y)=mini=s/2s/2j=s/2s/2IRx+i[y+j]IL[x+i+dv][y+j]

where

IR

and

IL

are pixel intensity value of right and left images and s means size of searching window.

Fig. 4. Object distance from camera [20].

Whereas in another paper suggested the object detection technique to be to detect the sudden global change on the lighting condition. This has to be done when the movement of the train is in OFF state. This lighting condition is analysed with the help of filtering technique in image processing [17]. This process is explained in block diagram below

Fig. 5. block diagram of object detection by image processing [17]

5.    FALSE ALARM FLITERING

Following steps are taken to eliminate false alarm [ 10, 17, 21]

•     Reflections on rails are localized on the tracks highness, enabling accurate 3D positioning. Lateral range for the two rails of each way is defined, so if a movement is only located in this range, it is labelled as a reflection. It works with all the cameras.

•     Gleams of the train on the opposite tracks are associated with the train movement. Moreover, they appear in the middle of the railway while drops and falls come from platforms.

•     Platform edge crossing. When a person stands close to the platform edge, his head may be over the tracks. The movement region is generally shared between the platform zone and the tracks zone. Event is only raised when the movement is disconnected from the platform or its surface over the tracks zone superior to the one on the platform zone. But clothes of the same colour as the background may prevent segmentation algorithm to connect it to the platform, and the system would consider it is a launched or dropped object. The rule for filtering this case is a fall (object dropping) or launch (over the tracks) trajectory condition.

•     Light effects may cause large variations in the image, such as headlamps gleams, halos, camera saturations. They are filtered with rules on maximum 3D sizes.

6.    NON CRITICAL ALARM

Proximity warning:  The proximity warnings are rather different depending on the point of view on each camera. When a traveller is close to the edge of the platform, the top of his body is over the white line. By the use of cameras, the location of the users’ feet image location (defined as the bottom of the extracted region) to tell if they are over the white line or not [10,17].

Object dropping warning: The object drops are qualified by the origin of the movement (e.g. the platform) and the fall trajectory (global vertical movement, top to bottom direction in the image) [10].

Object launching alarm: These events are qualified by their origin and destination, and by the global trajectory over the tracks. Since it is an alarm, the raising delay will be shortening as soon as the trajectory type can be sufficiently qualified as a launching [10]. 

7.    CRITICAL ALARM

These events impose to raise an alarm with the shortest delay to fit the electricity supply interruption need. Therefore, the analysis can only use the data taken from the first frames of the event. The main difficulty is to distinguish human going down on tracks from proximity warnings and object dropping, in order to prevent the critical alarm to raise in- stead of a simple warning. Few of the critical alarms are as mentioned below [21- 23]

Person trapped by door of a moving train: This situation is restricted to the cameras which are placed over the platform. This case seems similar to the proximity warning case while the train is moving. Il this case the motion is only detected on the platform zone of the dynamic mask, so only the movement on the bottom of the trapped body is detected (legs).

Walking on the rails:The conditions are the 3D size compatibility with human size and the disconnection from the platform zones.

Fall on tracks:This event category includes volunteer step down (slowly or jump) on the tracks, and accidental falls or pushes. The fall event is raised when the centre of gravity of the movement crosses the vertical separation of the plat- form edge.

Crossing the rails:The events starts after a fall alarm when the movement goes away for platform (condition on lateral distance to the edge of the platform), and stops when the person reaches the opposite platform.

8.    ADVANTAGES / MAIN LIMITATION

The metro environment has been deeply studied, and the main concealed difficulties have been identified. Most of them have been filtered with efficient dedicated rules [16,12]. Nevertheless, motion detection algorithms only extracts regions with connexity-based rules, with no more information than image and 3D extrapolated sizes and positions. Therefore, platform camera viewpoints show intrinsic limitations for the detection when the platform is crowded and when any dangerous situation appears.

The analyse of the dataset shows a lack of anticipation for alarm raising but the false positive rate for each event type remains very low. However, the metadata analysis can be tuned to adapt the reactivity of detection or false alarm rates according to the operational constraints.

9.    CONCLUSION

Various technologies are in use for railway platform safety all over the world. Precisely in Melbourne there is not any particular system to keep track of objects in railway track. In Japan, for platform safety automated glass barriers are used, where the barrier will be inserted in the platform once the train arrives on the platform. This technology sounds sophisticated but is very hard to implement as whole platform infrastructure needs to be changed will be very expensive. Vision based system is easy to implement and install and no further training of staffs is necessary. Automated glass technology needs constant human monitoring but vision based system is fully automated with minimal human interaction. This system monitors almost entire length of the track line in the platform, and determines in real-time whether human or dangerous obstacle is present by using image processing technology. Detection of train state and object is conducted robustly by using proposed image processing algorithm. Currently, a new recognition method is perused using stereo vision which calculates automatically volume of objects in monitoring area. Moreover, other dangerous factors are considered, such as safety accidents as fall between a platform and a train, getting stuck between the doors and disastrous fire etc. Moreover, to deal with the accident immediately, an effective information transmission system such as data transmission to control centre by using 4G network is being considered for dealing with the safety accidents. It is expected that the proposed system will play a key role for establishing highly intelligent monitoring system in railway.

References

 

  1. T.Jacks, 2017, ‘Every eight days someone dies on Victoria’s rail network’, The Age, 1 November, viewed 22 September 2018,  https://www.theage.com.au/national/victoria/every-eight-days-someone-dies-on-victorias-rail-network-20171031-gzbwd5.html
  2. T. W. S. Chow and S. Y. Cho, “Industrial neural vision system for underground railway station platform surveillance,” Adv. Eng. Inform. 12, 73-83 (2002).
  3. C. Davies, H. Y. Jia, and S. A. Velastin, “Crowd monitoring using image processing,” Electr. Comm. Eng. J. 7, 37-47 (1995).
  4. S. Reqazzoni, A. Tesei, and V. Murino, “A real-time vision system for crowding monitoring,” in Proc. Int. Conf. Industrial Electronics, Control, and Instrumentation (Maui, HI, USA, Nov. 1993), pp. 15-19.
  5. S. Y. Cho and T. W. S. Chow, “A fast neural learning vision system for crowd estimation at underground stations platform,” Neur. Proc. Lett. 10, 111-120 (1999).
  6. S. A. Velastin, B. A. Boghssian, and M. A. Vicencio-Silva, “A motion-based image processing system for detecting potentially dangerous situations in underground railway stations,” Transp. Res. C Emerg. Technol. 14, 96-113 (2006).
  7. S. Bouchafa, D. Aubert, and S. Bouzar, “Crowd motion estimation and motionless detection in subway corridors by image processing,” in Proc. IEEE Conf. Intelligent Trans- portation System (Boston, MA, USA, Nov. 1997), pp. 9-12.
  8. B. P. L. Lo and S. A Velastin, “Automatic congestion detection system for underground platforms,” in Proc. Int. Symp. Intelligent Multimedia, Video and Speech Processing (Hong Kong, China, Apr. 2001), pp. 158-161.
  9. C.Sacchi and C. S. Regazzoni, “A distributed surveillance system for detection of abandoned objects in unmanned railway environments,” IEEE Trans. Vehic. Technol. 49, 2013-2026 (2000).
  10. C.Seyve, “Metro railway security algorithms with real world experience adapted to the RATP dataset,” in Proc. IEEE Conf. Advanced Video and Signal Based Surveillance (Como, Italy, Sep. 2005), pp. 177-182.
  11. A.Cavallaro, “Event detection in underground stations using multiple heterogeneous surveillance cameras,” LNCS: Advanced in Visual Computing 3804, 535-542 (2005).
  12. M. Spirito, C. S. Regazzoni, and L. Marcenaro, “Automatic detection of dangerous events for underground surveillance,” in Proc. IEEE Conf. Advanced Video and Signal Based Surveillance (Como, Italy, Sep. 2005), pp. 195-200.
  13. J. Black, S. Velastin, and B. Boghossian, “A real time surveillance system for metropolitan railways,” in Proc. IEEE Conf. Advanced Video and Signal Based Surveillance (Como, Italy, Sep. 2005), pp. 189-194.
  14. S.Oh, S.Park, C.Lee, ‘Vision Based Platform Monitoring System for Railway Station Safety’, IEEE, 1-4244-1178-5/07, (2007).
  15. S.Oh, G.Kim, H.Lee,’A Monitoring System with Ubiquitious Sensors for Passenger Safety in Railway Platform’, The 7th International Conference on Power Electronics, October 22-26, 2007/ EXCO, Daegu, Korea (2007).
  16. T.W.S. Chow, Siu.Y.C, ‘Industrial neural vision system for underground railway station platform surveillance’, Elsevier Science Ltd, S 1474-0346(01)00002-7, (2002)
  17. Y.Park, D.Lee, ‘3D Vision-based Security Monitoring for Railroad Stations’, 2010, Journal of the Optical Society of Korea, Vol.14, No. 4, pp. 451-457, (December 2010)
  18. M. Rüder, N. Möhler, F. Ahmed, ‘An Obstacle Detection System for Automated Trains’, IEEE, 0-7803-7848-2/03, (2003)
  19. H.Niu, T.Hou, ‘Fast Detection Study of Foreign Object Intrusion on Railway Track ’, Archives of Transport, Volume 47, Issue 3, 2018
  20. S.C. Oh, G.D.Kim, W.T.Jeong, Y.Park, ‘Vision-based Object Detection for passenger’s Safety in Railway Platform’, International Conference on Control, Automation and Systems, October 14-17, 2008, Seoul, Korea.
  21. B.Delgado, K.Tahboud, E.J.Delp, ‘Automatic Detection of Abnormal Human Events on Train Platforms’, IEEE, 978-1-4799-4689-1/14, (2014)
  22. C.Carincotte, X.Naturel, M.Hick, J-M Odobez, J.Yao, A.Bastide, B.Corbucci, ‘Understanding metro station usage using closed circuit television cameras analysis’, IEEE Conference on Intelligent Transport Systems, October 2008, Beijing, China.
  23. A.  Bigdeli, B.C. Lowell, C.Sanderson, T.Shan, and S. Chen, ‘Vision processing in intelligent CCTV for mass transport security’, IEEE workshop on Signal Processing Applications for Public Security and Forensics, Washington D.C., USA(2007).

Cite This Work

To export a reference to this article please select a referencing stye below:

Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.

Related Services

View all

DMCA / Removal Request

If you are the original writer of this essay and no longer wish to have the essay published on the UK Essays website then please: