Limitations of an Intelligence CCTV System - Essay Example

Limitations of intelligence CCTV system Institution Name Date Table of Contents Table of Contents 2 Introduction 3 Background 4 Limitation of intelligence CCTV system 7 Technological Limitations 7 Limited opportunity for information fusion 7 Tracking is computationally-intensive 8 Sensor synchronisation and coordination 9 High bandwidth requirement 9 Functional Limitations 10 Differentiating behaviours 10 CCTV not necessarily reliable 11 Legal and Social limitations 12 Tendency for non-neutrality 12 Personal privacy and liberty concerns 13 Economic limitations 14 Prohibitive costs yet less effective 14 Human factor limitations 15 Informal practices and deviance of the human element 15 Conclusion 16 References 17 Introduction Rapid technological growth had led to increased network bandwidth and enhanced image processing, both of which have contributed to the realisation of intelligent video surveillance system. Analogue or traditional closed-circuit television (CCTV) require relatively few operators to monitor substantial number of cameras in protected areas, including airport, roads, military barracks or city centres (Hengstler et al., 2007). Intelligent surveillance systems have the capacity to offer automated services, including detection of sudden incursion, counting people and monitoring robbery. Generally, video surveillance systems are designed to monitor particular activities through analysis of recorded video or multi-channel monitoring. Examples of the intelligent surveillance system include the intelligent closed-circuit TV (ICCTV) (Saini et al. 2014; Chiemele et al., 2012). Intelligent closed-circuit TV use powerful computers to scrutinize and analyse video feeds in order to assist the human operators to detect actions of interest in real-time and to deter criminal activities and other undesired behaviours (Rayner, 2008). A proactive ICCTV has the capacity to process video streams on the fly by identifying objects or events of interest as they happen. Security technology evaluation is a critical process that makes up the key countermeasures for the protection of assets based on the identified security risks. Such countermeasures can be regarded as a combination of security equipment, procedures and policies (Smith, 2007). Therefore, evaluating security technology within the security environment of the protected facility contributed to a major function in treatment of the identified risks. Based on this premise, evaluation of intelligent CCTV system is critical. Several researchers have concluded that despite the evolution and promises enhanced surveillance, the intelligence CCTV is not very effective in uncontrolled environments, or outdoors, although it remains effective in controlled environments, such as indoors, because the adverse environmental effects are eliminated (Saini et al. 2014; Chiemele et al., 2012). This paper shows that Intelligent CCTV are not panacea to crime and disorder, since despite much excitement in the academia and security professionalism about their capabilities, technological, functional, legal, social, economic, and human factor limit their efficiency, effectiveness and applicability. Background The video surveillance system has evolved through three decades since it was initially used. The main drivers for evolution of the CCTV include the need for reduced costs, better image quantity and quality, remote-monitoring capabilities, greater limit in system intelligence, longer retention of captured videos and scalability and size (Bradley, 2003). To accommodate the requirements, video surveillance had undergone several technological shifts, the latest being from analogue CCTV to intelligence CCTV, with a full-fledged digital network-based video surveillance system (Bosch, 2014). Intelligent video surveillance system describes the process of monitoring events or activities in real-time, within a particular environment of a transient and a persistent object. The system’s core priorities entail maintaining access control to secured environments, monitoring the perimeter for intrusion and watching over protected personalities, unattended objects and suspicious behaviours (Nicolescu & Medioni, 2000). Adoption of intelligent closed-circuit TV (ICCTV) has caused a paradigm shift in video surveillance, from central-control surveillance system to distributed-control surveillance system (Chiemele et al. 2012). Intelligent CCTVs perform a combination of low-level image processing operations on the input frame, positioned at the edge of the sensor, which ensures video compression (Abidi et al., 2008). The processing and analysis of the video is mostly performed at the central host, through the use of standard workstation racks (See Figure 1). Figure 1: structure of video analytics (Chiemele et al. 2012). The Intelligent video surveillance technology is classifiable into distributed platforms and centralised platforms. In the centralised video system, analysis of the content of the video is executed at the back-end, through the use of DVR to help in processing and analysis of data from the cameras (Chiemele et al., 2012). However, in distributed platforms, intelligent network camera performs analysis of the video content in real-time, hence ensuring prompt response. The main advantages of using distributed platform include enabling easy expansion, addition of new cameras, decreased server workload, reduced cost of labour, accuracy of content and deployment costs. Figure 2: Centralised video content analysis Figure 3: Distributed content analysis Intelligent CCTV performs various computational applications. Initially, the applications were limited to image capture, compressing the image and transferring the image to the host computer. However, recently, intelligent CCTV perform a wide range of applications, with the most popular algorithms including real-time face recognition, scene capturing, traffic surveillance and human activity recognition (Kleihost et al., 2004; Ozer & Wolf , 2001; Matsushita et al., 2003; Bramberger et al., 2004; Bojkovic & Samcovic, 2006). Figure 4: Smart CCTV conceptualisation Limitation of intelligence CCTV system Technological Limitations Limited opportunity for information fusion To determine the limitations of intelligent CCTVs, Saini et al. (2014) investigated why the technology is rarely used in commercial systems. According to Saini et al. (2014), the most critical limitation of the technology is the limited opportunity for information fusion. The researchers echoed an earlier study by Atrey et al. (2006) that had established that during the video analysis process, information fusion only happens at three key levels, namely the data level, where the pixel values are compared directly to reach a conclusion, the feature level, where features are extracted from the image and the decision level, where conclusion is reached and acted upon (See Figure 5). Figure 5: Processing time of individual processes (Saini et al., 2014) Saini et al. (2014) concluded that the smart camera systems in the intelligent CCTV systems only allow fusion to take place at the decision level. It is submitted that this is since the multi-camera system consist of cameras that are densely positioned with overlapping views that need data and feature level fusion. In a related study of the limited opportunity for information fusion, Mitchell (2012) reached a similar conclusion. To overcome such limitations, Yang et al (2011) and Maker (2009) proposed compression feature techniques, which may however still compromise the general accuracy of the video analysis task. Tracking is computationally-intensive As indicated in the figure 5 above, the processing times of the several steps involved in video analysis system of four videos differs. The foreground detection is dependent on the frame resolution while the processing times of tracking and detection are relative to the computational load in each step (Atrey et al., 2006). Hence, tracking could be viewed to be computationally-intensive. Sensor synchronisation and coordination Sensor synchronisation and coordination may also be difficult in the intelligent CCTVS because of random network delays at the transitional nodes. This views was promoted by Farenbook and Clement (2011). Sensor synchronisation and coordination limitation has the potential to hinder the cameras from providing effective tracking performance as high-quality video from at one node without encouraging extra bandwidth overhead (Tessens et al., 2008). To overcome such limitations -- in order to achieve high level of tracking accuracy--, it is reasoned that the centralised system should have high quality video by integrating multiple cameras, which cause large bandwidth overhead. High bandwidth requirement The high bandwidth requirement is an underlying limitation to attaining a full-fledged system. This limitation was discussed by Bigdeli et al. (2008) when designing smart cameras that enable proactive intelligent CCTV. Bigdeli et al. (2008) investigated the possibility of designing cameras that could overcome the high bandwidth requirement that high resolution cameras need, by discarding unutilised information at the sensor, before the image of interest is moved to the CPU. As a result setting aside CPU time needed for actual processing. The researchers found that for accuracy and efficiency in recognition, high-resolution camera is required which in turn needs a large bandwidth to transfer the image to a central processing unit (CPU). The resultant memory management and data handling in actual sense paralyses the CPU. In turn, no processing power is left to process the image and execute the recognition (Wilson, 2003). Additionally, to determine the best views, the smart cameras in the intelligent CCTV have to share foreground information. The figure below indicates the proportion of the image area for the foreground of 24-hour real surveillance footage. As indicated, the foreground area may range from zero to about 63 percent. Figure 6: mount of foreground in real surveillance footage of 24 hours (Saini et al. 2014). Therefore, sharing such a large amount of data requires a huge amount of bandwidth. Additionally, in overlapping camera, best views can alter between consecutive frames, which would need frequent shifts in the role of the master camera. Changing the master camera repeatedly would require extra bandwidth, as well as processing overhead. To overcome such situations, Saini et al. (2014) suggested that normal IP cameras could be used to capture video and assign all video analysis to private cloud. Functional Limitations Differentiating behaviours The intelligent CCTVS face difficulty in differentiating suspicious from usual behaviours of events and objects. This implies that the likelihood of false alarms is imminent, especially where the intensions of the object are legitimate, including when an individual is running to halt a cab across the street, or where the object’s speed is faster than the threshold, configured in the video surveillance system (See Figure 7) (Nam et al. 2010). Figure 7: intelligent surveillance system appropriated services (Nam et al. 2010) At any rate, if the threshold speed is increased, a likelihood of a real threat going unobserved is also increases. Velastin (2009) also discussed that the requirement of low false alarms and high detection rates are major challenges. Environmental changes in the image’s background may result to errors in video analytics and analysis of threats. Such environmental changes comprise moving lighting effects from the clouds or the sun, which may cause shadows. CCTV not necessarily reliable Intelligent CCTV may not often be able to monitor each section of the protected area, such as building or street. Despite the fact that the cameras are positioned strategically, there is limited guarantee that they will capture all unwarranted behaviours or criminal activity. This could be due to the likelihood of dust enveloping the lens screen or spray painting by the perpetrators. Vandalism during street riots also mean that the cameras may be destroyed, hence limiting their capacity to identify perpetrators of a crime or monitoring criminal activities (Wolf et al. 2002). Hence, not all actions are likely to be recorded, which limits the capacity of this technology to deter crime. Legal and Social limitations Tendency for non-neutrality Despite the limited empirical work on use of intelligent CCTV to monitor certain subject populations, the evidence thus far indicates that disproportionate targeting of groups or individuals that are perceived to be out of place in town centres or streets, such as the homeless, youngsters, junkies and other categories of 'undesirable individuals, may contribute to disproportionate reporting (Dubbeld, 2003). In essence, this is since the cameras may be configured to identify certain body features, movements and behaviours that are viewed to indicate the potential of crime in order to construct and identify these targeted populations. According to Dubbeld (2003) such tendency appears to be stronger with intelligent CCTV that applies automated face recognition systems. Fatemi et al (2003) stated that the algorithmic software designed for such purposes may be configured to unequivocally focus on the body, since its logic is anchored entirely in analysing the facial physiognomy of discrete individuals. This implies that the intelligent cameras are inherently partial or non-neutral. In view of this, it could be argued that the intelligent CCTV cameras are likely to miss out on the real perpetrators of crime. Additionally, such non-neutral tendencies have the potential to raise ethical and legal issues (Brooks, 2001). This perspective is supported by Dubbeld (2003). In his view, asymmetrical relations exist between those being watched and the watchers. The asymmetries entail visibility, embodiment and knowledge. According to Smith (2004), there is fear that intelligent CCTV may not only be used for criminal surveillance but also monitoring social discredited persons or groups, who are viewed to be threats to economic prosperity of an area, such as gangs of youths or beggars. Personal privacy and liberty concerns Fletcher (2011) described CCTV as being both benevolent and malevolent entity that continuously modernises inconsistency with modernisation of society. According to the researcher, Britain, which is the country with most intelligent cameras on the streets, airports, stations and city centres had some 5 million CCTV cameras, estimated at one for every 12 people. The large number of CCTV cameras has called into question the freedom of privacy. In Fletcher’s view, the real issue has been whether they are actually to protect the public or to promote the concept of "Big Brother" surveillance society. The Intelligent CCTV monitors an individual's walking style and the facial physiognomy to those stored in a database, hence preventing perpetrators from hiding from CCTV. According to Fletcher (2011), while intelligent CCTV signifies crucial technological advancement, it has vast limitations. Fletcher (2011) conducted a review of trials of the system and observed that the accuracy rate of the facial recognition may not be reliable, exposing some people to harassment by the police in situations where the software wrongly identifies people. He suggested responsible use of the system. In which case, although intelligent CCTV can be used for benevolent purposes of promoting security to the public, they have been deployed reservedly due to concerns about personal privacy and liberty. Economic limitations Prohibitive costs yet less effective Velastin (2009) pointed out that despite the increased number of intelligent CCTVs in private and public places globally, the cost of monitoring the cameras and analysing recorded video is greatly prohibitive, hence a major limitation to adoption and application of intelligent CCTVs. Nam et al. (2010) shared a similar perspective in suggesting that resourcing human operators to monitor the intelligent CCTV is prohibitively expensive. Additionally, there is a tendency to concentrate the monitoring in extensive control rooms. For larger systems, such as in airports, city centre of metropolitan railways, thousands of cameras may be required, which indicate high procurement, installation, maintenance and monitoring costs. Despite this, only a small fraction of installed cameras is watched. The perception is supported by Velastin (2009) who studied the railway sector in UK and established camera to screen ratio of between 1:4 and 1:78. This shows that only a small fraction of the cameras can be watched in real-time. In other cases, they may be driven reactively from reports from people on the ground. From this, it is reasoned that despite the high cost incurred in procuring thousands of cameras to enhance security of a place, only a few can be viewed in real time at any one time. This also means that the capacity of the intelligent cameras to deter crime, such as terrorism, may be exaggerated as not all cameras can be watched at any one particular time. Human factor limitations Informal practices and deviance of the human element Smith’s (2004) study of informal practices and deviance among CCTV control room operators in the UK also supported the assumption that Intelligent CCTV is not panacea to crime and disorder. Indeed, much of the literature appears to have a form of technological determinism, where the human element operates the cameras in the control rooms and their role in the effectiveness or the efficiency of the CCTV cameras is ignored. Smith (2004) outlined that human element such as irrationality, resistivity, prejudicial and dysfunctional limit the potential of the cameras in detecting and deterring crime in protected areas. Basing on Smith’s (2004) argument, it could be argued that most empirical studies on the limitations of CCTV cameras have overlooked the fact that the cameras are neither autonomous nor conscious. On the contrary, for the camera’s to be effective in their primary objective of detecting and deterring crime human beings have to constantly control and monitor them in work-like situations, so as to ensure that millions of the images that are generated are observed, interpreted and executed. In fact, without the three – observation, interpretation and response – installation of the intelligent cameras would be futile. Drawing on Smith’s (2004) ethnographic study he conducted in a CCTV control room, it is further argued that intelligent CCTV cameras are susceptible to the human factor. In which case, failure to control the control rooms effectively limits their effectiveness. Smith (2004) found that CCTV control rooms are not controlled effectively, as it is rife with time wasting, playfulness, comedy and reading magazines even as the operators sought to address socio-structural disparities, as well as the monotony of unrewarding and routine jobs. He further argued that the human elements assigned the duty of operating the CCTVS are subjective human beings who are susceptive to emotional and natural factors, including frustration, daydreaming, boredom, fatigue, discrimination and apathy. Such factors limit the effectiveness of the CCTV cameras in detecting and deterring crime in real-time. Under such circumstances, it is suggested that employing more people to operate the cameras and having work routines throughout the day can effectively ensure efficiency of the human element. Conclusion Intelligent CCTV are not panacea to crime and disorder since, despite much excitement in the academia and security professionalism about their capabilities, technological, functional, legal, social, economic, and human factor limit their efficiency, effectiveness and applicability., effectiveness and applicability. Technological limitations include limited opportunity for information fusion since the smart camera systems in the intelligent CCTV systems only allow fusion to take place at the decision level. Tracking is also computationally-intensive since foreground detection is dependent on the frame resolution while the processing times of tracking and detection are relative to the computational load in each step. Sensor synchronisation and coordination may also be difficult in the intelligent CCTVS because of random network delays at the transitional nodes. They require high bandwidth as their smart cameras have to share foreground information. Functional limitations include the fact that the intelligent CCTVS face difficulty differentiating suspicious from usual behaviours of events and objects. They may also not often be able to monitor each section of the protected area, such as building or street. They have the potential to violate privacy and rights. Although intelligent CCTV can be used for benevolent purposes of promoting security to the public, they have been deployed reservedly due to concerns about personal privacy and liberty. Lastly, despite the increased number of intelligent CCTVs in private and public places globally, the cost of monitoring the cameras and analysing recorded video is greatly prohibitive, hence a major limitation to adoption and application of intelligent CCTVs. References Abidi, B. NAragam, N., Yao, Y. & Abidi, M. (2008). Survey and analysis of multimodal sensor planning and integration for wide area surveillance. ACM Computing Surveys, 41(1). article 7 Atrey, K., Kankanhalli, M. and Jain, R. (2006). Information assimilation framework for event detection in multimedia surveillance systems. Multimedia Systems, 12(3), 239–253 Bigdeli, A., Lovell, B. & Shan, T. (2008). Smart Camera: Enabling Technology for Proactive Intelligent CCTV. Retrieved: Bojkovic, Z. & Samcovic, A. (2006). Face Detection in Neural Network based method for video surveillance. Neural Network Application in Electrical engineering Bosch. (2014). Adding Intelligence to an Existing CCTV System Solution Brief. Bosch Security Systems, Inc Bradley, A. (2003). Can Region of Interest Coding Improve Overall Perceived Image Quality? In proceedings of WDIC2003, Feb 2003: 41-44. Bramberger, M., Brunner, J., Rinner, B., Schwabach, H. (2004). Real-Time Video Analysis on an Embedded Smart Camera for Traffic Surveillance. In proceedings of the 10th IEEE Real-Time and Embedded Technology and Applications Symposium (RTAS’04), May 2004: 174- 181. Brooks, D.J. (2001). Closed Circuit Television: Legal considerations for the security industry regarding digital processed video images. In H.Armstrong (Ed.). Proceeding of the 5th Australian Security Research Symposium. Perth: Edith Cowan University. pp. 29-42 Chiemele, A., Olushola, I. & Onyinyechi, N. (2012). Intelligent Video Surveillance System Case Study On Huawei’s Intelligent Video Surveillance System: London: London Metropolitan University Dubbeld, L. (2003). Observing bodies. Camera surveillance and significance of the body. Ethics and Information Technology 5(1), 151-162 Farenbook, J. & Clement, A. (2011). Hidden Changes: from CCTV to “Smart” video surveillance. Retrieved: Fatemi, H., Kleihorst, R., Corporall, H., & Jonker, P. (2003). Real-time Face Recognition on Smart Camera. Proceedings of Acivs 2003 (Advance Concepts for Intelligent Vision Systems), Belgium, September 2003: 222-227. Fletcher, P. (2011). IS CCTV Effective in Reducing Anti-Social Behaviour. Internet of Criminology 1-41 Hengstler, D. Prashanth, S. Fong, & Aghajan, H. (2007). Mesheye: a hybrid-resolution smart camera mote for applications in distributed intelligent surveillance. In Proceedings of the 6th International Symposium on Information Processing in Sensor Networks (IPSN '07), pp. 360–369, Kleihorst, R., Reuvers, M., Krose, B. & Broers, H. (2004). A Smart Camera For Face Recognition. Proceedings of International Conference on Image Processing (ICIP 2004): 2849-2852. Makar, M., Chang, D. Chen, S. S. Tsai, & Girod, B. (2009). Compression of image patches for local feature extraction,” in Acoustics. In IEEE International Conference on Speech and Signal Processing (ICASSP '09), pp. 821–824, IEEE, 2009. Matsushita, N., Hihara, D., Ushiro, T., Yoshimura, S., Remkimito, J. & Yamamoto, Y. (2003). ID CAM: A Smart Camera for Scene Capturing and ID Recognition. In proceedings of the Second IEEE and ACM International Symposium on Mixed and Augmented Reality (ISMAR ’03). October 2003: 227- 236. Mitchell, H. (2012). Introduction in Data Fusion: Concepts and Ideas. Springer, pp. 1–14, Nam, Y., Rho, S. & Park, H. (2010). Intelligent video surveillance system: 3-tier context-aware surveillance system with metadata. New York: Springer Nicolescu, N. & Medioni, G. (2000). Electronic Pan- Tilt-Zoom: A Solution for Intelligent Room Systems. In proceedings of IEEE International Conference of Multimedia and Expo (ICME 00), IEEE CS Press, Los Alamitos Calif., 2000: 1581-1584. Ozer, B. & Wolf, W. 2001. A smart camera for real-time human activity recognition. In Proceedings, SIPS-01, IEEE: 217-225. Rayner, G. (2008). New intelligent CCTV cameras can see and hear. Retrieved from The Telegraph website: Saini, M., Atrey, P. El Saddik, A. (2014). From Smart Camera to SmartHub: Embracing Cloud for Video Surveillance. International Journal of Distributed Sensor Networks Volume 2014 (2014), Article ID 757845, 10 page Smith, G. (2004). Behind the Screens: Examining Constructions of Deviance and Informal Practices among CCTV Control Room Operators in the UK. Surveillance & Society 2(2/3), 376-395 Smith, C. (2007). The Evaluation of Security Systems: Testing Biometric and Intelligent Imaging Systems. Keynote Address: The Sixth International Workshop for Applied PKC (IWAAP2007) Tessens, L., Morbee, M., Lee, H. Philips, W., & Aghajan, H. (2008).Principal view determination for camera selection in distributed smart camera networks,” in Proceedings of the 2nd ACM/IEEE International Conference on Distributed Smart Cameras (ICDSC ’08), September 2008 Velastin, S. (2009). CCTV Video Analytics: Recent Advances and Limitations. Berlin: Springer-Verlag 22-34 Wolf, W., Ozer, B. & Lv, T. (2002). Smart Cameras as Embedded Systems. IEEE Computer, 35(9), 48–53. Wilson, A. (2003). Smart Cameras embed processor power. Vision Systems Design, 95-99. Yang, A., Maji, S.. Christoudias, C. Darrell, T., Malik, J. & Sastry, S. (2011). Multiple-view object recognition in smart camera networks, in Distributed Video Sensor Networks, pp. 55–68, Springer, 2011 Read More

Figure 2: Centralised video content analysis Figure 3: Distributed content analysis Intelligent CCTV performs various computational applications. Initially, the applications were limited to image capture, compressing the image and transferring the image to the host computer. However, recently, intelligent CCTV perform a wide range of applications, with the most popular algorithms including real-time face recognition, scene capturing, traffic surveillance and human activity recognition (Kleihost et al.

, 2004; Ozer & Wolf , 2001; Matsushita et al., 2003; Bramberger et al., 2004; Bojkovic & Samcovic, 2006). Figure 4: Smart CCTV conceptualisation Limitation of intelligence CCTV system Technological Limitations Limited opportunity for information fusion To determine the limitations of intelligent CCTVs, Saini et al. (2014) investigated why the technology is rarely used in commercial systems. According to Saini et al. (2014), the most critical limitation of the technology is the limited opportunity for information fusion.

The researchers echoed an earlier study by Atrey et al. (2006) that had established that during the video analysis process, information fusion only happens at three key levels, namely the data level, where the pixel values are compared directly to reach a conclusion, the feature level, where features are extracted from the image and the decision level, where conclusion is reached and acted upon (See Figure 5). Figure 5: Processing time of individual processes (Saini et al., 2014) Saini et al. (2014) concluded that the smart camera systems in the intelligent CCTV systems only allow fusion to take place at the decision level.

It is submitted that this is since the multi-camera system consist of cameras that are densely positioned with overlapping views that need data and feature level fusion. In a related study of the limited opportunity for information fusion, Mitchell (2012) reached a similar conclusion. To overcome such limitations, Yang et al (2011) and Maker (2009) proposed compression feature techniques, which may however still compromise the general accuracy of the video analysis task. Tracking is computationally-intensive As indicated in the figure 5 above, the processing times of the several steps involved in video analysis system of four videos differs.

The foreground detection is dependent on the frame resolution while the processing times of tracking and detection are relative to the computational load in each step (Atrey et al., 2006). Hence, tracking could be viewed to be computationally-intensive. Sensor synchronisation and coordination Sensor synchronisation and coordination may also be difficult in the intelligent CCTVS because of random network delays at the transitional nodes. This views was promoted by Farenbook and Clement (2011).

Sensor synchronisation and coordination limitation has the potential to hinder the cameras from providing effective tracking performance as high-quality video from at one node without encouraging extra bandwidth overhead (Tessens et al., 2008). To overcome such limitations -- in order to achieve high level of tracking accuracy--, it is reasoned that the centralised system should have high quality video by integrating multiple cameras, which cause large bandwidth overhead. High bandwidth requirement The high bandwidth requirement is an underlying limitation to attaining a full-fledged system.

This limitation was discussed by Bigdeli et al. (2008) when designing smart cameras that enable proactive intelligent CCTV. Bigdeli et al. (2008) investigated the possibility of designing cameras that could overcome the high bandwidth requirement that high resolution cameras need, by discarding unutilised information at the sensor, before the image of interest is moved to the CPU. As a result setting aside CPU time needed for actual processing. The researchers found that for accuracy and efficiency in recognition, high-resolution camera is required which in turn needs a large bandwidth to transfer the image to a central processing unit (CPU).

Read More