Caught on camera, lost in transmission: CCTV lapses reveal hurdles in solving crime

Only 5,100 of the 7,500 cams are operational.

The recent blast at The Rameshwaram Cafe in the eastern part of Bengaluru City has once again sparked a debate about the efficacy of CCTV cameras in the city. This is particularly relevant given that the rates of solving many public crimes—which rely heavily on CCTV footage—have remained below the 50% mark over the past few years.

A little over a month ago, CCTV cameras played a crucial role in locating a 12-year-old school student who went missing in the city. Now, footage from these cameras is proving invaluable in assisting the police—including both the local Central Crime Branch (CCB) and the National Investigation Agency (NIA)—in tracking the suspect involved in the low-intensity blast that occurred at The Rameshwaram Cafe in Brookefield on March 1. However, there are several challenges.

Safe City Project

Under the Safe City Project, which is estimated to have cost Rs 661.5 crore, the police have adopted “cutting-edge” technology and installed 7,500 CCTV cameras across the city. In addition, nearly 3 lakh cameras have been installed under the Karnataka Public Safety (Measures) Enforcement Act, 2017.

However, police sources indicate that only 5,100 of the 7,500 cameras are operational. The connectivity for the remaining cameras has been delayed due to prolonged issues encountered with their telecommunication partner.

The department has scheduled a meeting with the Chief Secretary to the Karnataka Government on March 12 to discuss and address this issue. Sources say they are prepared to sever ties with their current telecommunication partner.

Street crimes

Dysfunctional cameras hamper police operations, such as the detection and tracking of suspects in public crimes or violations. Law and order police typically rely on CCTV footage to solve cases of motor vehicle theft, residential burglary, housebreaking, chain snatching, and other thefts.

According to data released by the City Crime Record Bureau, the detection rates for these crimes have mostly been below 60% since 2021, except for chain snatching cases, which have seen a nearly 90% detection rate.

In 2023, the number of registered cases surged compared to previous years, while the detection rate significantly declined, prompting various questions. However, police have attributed this trend to the frequent reassignment of officers and their election duty commitments.

Tracking in real time?

The police department recently announced that it has adopted ‘proactive’ and ‘preventive’ policing methods to leverage the AI-backed technology at their disposal. These methods aim to track any individual in real time, provided they are under the surveillance of AI-equipped cameras.

During the ongoing blast investigations, an officer familiar with the case disclosed that the suspect’s movements within the city were primarily tracked using the available CCTV footage.

Furthermore, DH learned from a senior police department source that approximately 450-500 cameras were installed under the Safe City project in the vicinity of the blast site. Notably, one such camera, located just a few meters from the cafe, was found to be non-functional.

“He travelled mostly using public transport. Whenever he was out on the ground, most of his movements were learnt using CCTV footage,” claimed the officer.

This was despite the unavailability of footage in certain places. However, he refused to answer if they used solely CCTV footage to follow his trail claiming it as “part of the investigation”. The reason for the delay in tracing the suspect’s travel route, despite the availability of AI cameras and advanced technology, remains unanswered, as the police have not responded to media inquiries.

Limitations in AI

The cameras installed under the safe city project use facial recognition to track the person’s real-time movement. But in this case, the suspect was wearing a mask and a hat through most of the footage.

The 24/7 surveillance is managed by a software, Command Control Communications and Computer 4 intelligence (C4i), connected to a central Command Control Centre, all equipped with AI, according to a police source.

The criminal database is fed to the C4i software which enables tracking of a person, whose data is already available on the database. Unavailability of the suspect’s record in this case might be another reason.

AI in traffic

Traffic police have recognised that AI cameras are not infallible in detecting road violations. While the detection accuracy for the helmet rule is about 97% other violations such as signal jumping exhibit only about 75% to 80% accuracy leaving room for numerous errors to occur. To address this issue Traffic Police Chief M N Anucheth has deployed teams to re-examine older violations and validate each new violation flagged by the cameras. “This is an attempt to raise the road safety standards by making the system more transparent and less prone to errors” Anucheth said acknowledging that this process is labour-intensive.