Artificial intelligence (AI) has become increasingly important in the field of video surveillance. AI can be used to automate the analysis of video footage, which can help detect and identify objects, people, and events in real-time.
One application of AI in video surveillance is object recognition. With AI algorithms, surveillance cameras can detect and recognize objects, such as cars or people, and classify them into categories such as vehicles or pedestrians. This can help identify potential threats or suspicious activities in real-time, allowing security personnel to take appropriate action.
Another application is behavior analysis, where AI algorithms can analyze patterns of movement and detect anomalous behavior, such as people moving in the wrong direction or loitering in a restricted area. This can help security personnel identify potential security risks and respond to them more quickly.
AI can also be used for facial recognition, which can be particularly useful in identifying individuals who are known to be a security risk. By analyzing live video footage or archived video, AI algorithms can identify individuals and compare them to watchlists or databases of known offenders.
Overall, the use of AI in video surveillance can improve the effectiveness and efficiency of security operations, allowing for more timely and accurate responses to potential threats. However, it is important to balance the potential benefits of AI with privacy concerns and ensure that surveillance systems are used responsibly and ethically.
History
Statement of the problem
Humans are limited in their ability to vigilantly monitor live video surveillance footage, which has led to the demand for artificial intelligence that can perform the task more effectively. Due to the high number of cameras and the generally uneventful nature of many camera views, human operators quickly lose attention and focus, rendering their monitoring ineffective. As a result, video surveillance systems have typically been used for forensic purposes rather than for real-time monitoring. Wide-angle camera views have also been found to be insufficient for identifying perpetrators due to their low resolution. So, the need for artificial intelligence to overcome the limitations of human monitoring and make video surveillance more effective for security and surveillance purposes.
Earlier attempts at solutions
Humans are limited in their ability to vigilantly monitor live video surveillance footage, which has led to the demand for artificial intelligence that can perform the task more effectively. Due to the high number of cameras and the generally uneventful nature of many camera views, human operators quickly lose attention and focus, rendering their monitoring ineffective. As a result, video surveillance systems have typically been used for forensic purposes rather than for real-time monitoring. Wide-angle camera views have also been found to be insufficient for identifying perpetrators due to their low resolution. So, the need for artificial intelligence to overcome the limitations of human monitoring and make video surveillance more effective for security and surveillance purposes.
Earlier attempts at solutions
Motion detectors to cameras was initially seen as a solution to the shortcomings of human guards in watching surveillance monitors for extended periods of time. The idea was that motion detection would alert remote monitoring officers of potential intruders or perpetrators, eliminating the need for constant human vigilance. However, in outdoor environments, the constant motion of elements like leaves, litter, insects, and shadows caused numerous false alerts, making this solution impractical for outdoor surveillance. The statement suggests that motion detection may still be effective in indoor environments during non-operating hours
Advanced video motion detection
The next evolution in video motion detection reduced false alerts to a certain extent, but at the cost of complicated and time-consuming manual calibration. This approach detects changes in a target, such as a person or vehicle, relative to a fixed background. However, changes in the background, such as seasonal changes or other alterations, can lead to a deterioration in reliability over time. The statement suggests that responding to too many false alerts proved to be an obstacle, rendering this solution insufficient.
Advent of true video analytics
True video analytics, powered by machine learning and computer vision algorithms, can accurately detect and classify objects and movements in real-time. This technology can distinguish humans, vehicles, and other selected objects from the general movement of other objects and visual static or changes in pixels on the monitor. It works by recognizing patterns and identifying anomalies based on pre-defined rules or parameters. For example, it can detect if the number of people in a pre-defined area exceeds a certain limit during a defined time interval and send an alert if a violation occurs. The alert may include a red rectangle or “bounding box” around the detected object or intruder, and a short video clip of the incident. This technology has revolutionized the video surveillance industry by enabling proactive monitoring and intervention, rather than just post-incident investigation.
Practical Application
Real-time preventative action
The limitations of traditional video surveillance, such as low resolution, poor visibility in adverse conditions, and human error, have led to the development of artificial intelligence (AI) for video analytics. This AI technology can distinguish between the movement of humans and other objects, and can detect intruders with high reliability and low false alert rates, even in adverse conditions. The AI is trained using statistical models based on a large number of images of humans in various positions and postures. This technology has the potential to overcome the limitations of traditional video surveillance
Talk-down
The use of artificial intelligence in video surveillance has significant advantages over human monitoring alone. The limitations of traditional video surveillance systems are discussed, such as their inability to effectively detect intruders in adverse conditions, their reliance on human operators who may suffer from attention fatigue, and their impracticality for facilities with many cameras. The passage goes on to explain how true video analytics can overcome these limitations by distinguishing specific objects or patterns of movement from the general motion of other objects. One of the most significant advantages of this technology is its ability to send alerts in real-time to human officers or operators, who can then deter the intruder using public address systems.
Verified breach report
video analytics technology is a more effective and reliable way to detect intruders and crimes in progress compared to traditional methods such as human surveillance or burglar alarms. The technology can overcome limitations such as low frame rates, poor lighting conditions, and human attention span. Additionally, the system allows for immediate response and deterrence through the use of outdoor loudspeakers and direct communication with law enforcement. The passage also highlights the issue of false alarms from burglar alarms and the low priority response they receive from the police.
Behavioural analytics
Active environments
While rule-based video analytics can work well for some security applications, it cannot work in many active environments where hundreds or thousands of people belong all over the place all the time, such as a college campus, an active factory, or a hospital. In these situations, it is not possible to set rules that would distinguish between legitimate people and criminals or wrong-doers.
Overcoming the problem of active environments
Behavioral analytics is a self-learning, non-rule-based A.I. that classifies objects and events seen in video cameras. The system recognizes patterns in human behavior and identifies anomalies that deviate from the norm. Unlike rule-based analytics, which work mainly to detect intruders into areas where no one is normally present at defined times of day, behavioral analytics work where people are active to detect things that are out of the ordinary. It can detect unusual events such as a fire breaking out, vehicles driving the wrong way, or someone being thrown to the ground by an attacker, and alert security personnel. However, the system is situation-specific and would learn what is normal and abnormal for each environment it is monitoring.
What the artificial intelligence 'understands'
Artificial intelligence (AI) systems can recognize and classify objects based on their physical properties, such as size, shape, and motion, and learn typical patterns of behavior associated with those objects. The AI system does not have an inherent understanding of what a human, fire, or vehicle is, but rather identifies and classifies them based on their observable features.
The AI system then uses this knowledge of typical behavior to identify anomalies or unusual patterns that may indicate a potential threat or risk. For example, if the AI system has learned that humans typically walk on sidewalks and not climb up buildings, it may trigger an alert if it detects someone scaling the side of a building.
Overall, this statement highlights the importance of training AI systems to recognize and respond to anomalous behavior, which can help improve their accuracy and effectiveness in various applications such as surveillance, security, and safety.
Varies from traditional mindset of security systems
The limitations of traditional alarm systems and the potential benefits of behavioral analytics in detecting anomalous behavior. While burglar alarms are effective at detecting true positives, they often have a high rate of false alarms. Motion detecting cameras can also miss true positives and have a high rate of false alarms, especially in outdoor environments. Rule-based analytics can detect true positives with low false positives but are limited to detecting the presence or absence of an intruder and cannot detect more complex events.
Behavioral analytics, on the other hand, can detect more subtle events such as fights breaking out or employees breaking safety procedures. While it may also trigger false alarms, the degree of deviation from normal behavior can be set to minimize these instances. This new way of human and AI interaction requires a shift in mindset from the traditional alarm industry, where false alarms are seen as a major issue. Instead, the AI acts as a “tap on the shoulder” to alert human officers to potentially anomalous behavior that may require a response.
Overall, the use of behavioral analytics represents a more sophisticated and nuanced approach to detecting threats and risks, which can improve the effectiveness of security systems and help prevent incidents from occurring.
Limitations of behavioral analytics
Due to the continuous processing of a large number of complex data, the software utilizes a lower resolution of only 1 CIF to reduce computational demand. However, this lower resolution may not detect objects that are too small or too far away, such as humans beyond a distance of sixty to eighty feet depending on conditions, if the camera used is wide angle. On the other hand, larger objects like vehicles or smoke can still be detected at greater distances.
Quantification of situational awareness
the practical use of artificial intelligence in security and its economic importance. It explains that A.I. is becoming increasingly utilized for security purposes, and also for non-security applications like operational efficiency and shopper heat-mapping. The article notes that A.I. can handle large data sets and scanning beyond human capability, but humans remain superior in judging subtleties of behavior or intentions. The article further explains that companies typically spend much less on security than the actual losses they incur, due to cognitive dissonance and underestimation of consequential losses. Finally, the article discusses the potential of A.I. in the form of behavioral analytics to proactively prevent incidents and reduce costs for employers.