On 16th January 2025, a Bollywood couple in Mumbai had an intruder in their 12th floor apartment. In a squabble that ended in a knife attack, the Bollywood star was gravely injured with six stab wounds, including two deep cuts in the back out of which one close to the spine could have been fatal. At midnight 2.30 am when the crime transpired, the star was rushed to the hospital where it was found that part of the knife was lodged near the thoracic spine. A neurosurgery was needed to take out the knife. Meanwhile, the Mumbai police examined CCTV footage of the building but couldn’t trace the attacker leading to an initial opinion that it might be an inside job. They put out the prima facie assessment that attacker didn’t enter the building within 2 hours prior to the incident. The public were puzzled when news reports emerged that attacker hid for 2 hours inside the building because CCTV didn’t show entry of the suspect before the attack. Later, police debunked the theory stating that accused was found in another CCTV and had taken the fire escape staircase to enter the actor’s house. What happened in the intervening period when police wasn’t able to locate the accused in the cameras? What took so long for locating the attacker? Didn’t the time spent in going through the CCTV’s give an advantage to the accused in escaping the law?
How video analytics augment the surveillance network?
On further investigation by the police, it was observed that attacker had entered the flat of the victim with a criminal intent. He entered the building compound by climbing a small wall from an adjacent building before taking the staircase. The police arriving at the crime scene next morning collected the CCTV visuals and went about examining the same. While details of the investigation aren’t yet public, it’s quite likely that CCTV outside the lift was checked first. A team of officers observed the activity by rewinding the recording by a few hours, verifying the identity with the guards stationed outside the gate and the residents of the society. After prolonged inspection, they didn’t find the attacker entering the lift and ended up with the wrong conclusion that attacker was already inside. The attacker may have preempted it because of which he chose the tiring staircase to reach the 12th floor instead of the easier option of taking the lift.
The wrongful dissemination of information to media was because police took a lot of time in checking the remaining surveillance cameras. In any crime, the chances of finding and catching the offender decreases with increase in time. When police take a lot of time to find the attacker and capture a photo for the purpose of tracing, it gives valuable hours to criminals. When it happens in a high-profile neighborhood in a city like Mumbai, there is panic especially among middle-class neighborhoods’ where security is relatively lax. With the advent of AI-powered video analytics softwares, the process of analyzing video footage can be automated. The wrong statement of Mumbai police claiming that no one was seen on CCTV was a by-product of time constraints before thoroughly analyzing all the footages while in a hurry to take quick action. The officers succumbed to media pressure as public opinion and political attacks were mounting. Examining a 5 hour video footage will naturally take 5 hours for a human observer. It gets more time when a high-security building has multiple CCTV’s from various angles such that no corner is outside their surveillance. When you have a high-profile case with government agencies breathing through your neck, human errors of judgment are likely.
How video analytics work?
AI video analytics leverages algorithms and machine learning to interpret video data automatically, facilitating real-time processing and decision-making. These AI systems can perform tasks such as object recognition, event detection, and anomaly identification with high accuracy. Apart from speeding up routine surveillance tasks, AI video analytics are highly adaptable in diverse settings like retail, healthcare and security sector. Integration of video analytics with IOT devices and smart infrastructure enhances capability of these systems to process video, real-time threat-detection and enable timely human response. Manual monitoring of surveillance systems can result in cost savings and ability to prevent incidents by alerting the human security staff. In the incident of stabbing of Bollywood star, the AI video analytics could have prevented the incident by alerting the guards of suspicious activity when the intruder was climbing the adjacent wall. A real-time threat appraisal of that event would have neutralized the threat before the attacker reached inside the flat of the actor.
Conclusion:
In an era where urban neighborhoods are equipped with advanced security systems like CCTV cameras, thumb-print access and face scanning technology, the security agencies and law enforcement officials are burdened by excessive digital data. Scouring through hours of records or hundreds of samples to find the 15 seconds of evidence isn’t a viable option when dealing with immediate threats or serious threats like terrorism. The precious time spent in sifting through terabytes of data can be reduced if video analytics applications assist humans in quickly zeroing in on the threat. In the Bollywood star stabbing incident, the first visuals of attacker was located and released by the police only in the late evening of the next day. Needless to say, it provided a wide window of timespan for the accused to seek a hideout or even leave the country before the image was circulated among the general public. With AI video analytics, the response-time could be reduced so that neighborhoods are secure and actionable inputs to the human guards and police are provided within minutes of the incident.
