BENGALURU: Recognising the need to have cutting-edge technology on their side in the fight against crime, the Bangalore Police, for the past three months, has been working on a project which would enable them to use closed-circuit television (CCTV) cameras to identify wanted persons as well as stolen vehicles.
So far, with a total of 731 cameras, this technology has been used to detect 37 stolen vehicles in the last one week. However, last Thursday, the cameras with facial recognition technology were used for the first time to identify two pickpockets who were spotted at the Yeswanthpur Railway Station. But while the technology performed as per expectations, the robbers managed to flee the spot as the police did not reach there in time to nab them.
“It is difficult for Hoysalas (car patrols) to catch them as they may be positioned in one direction while the suspects are heading in the opposite. To deal with this, we have now deputed constables in these areas so that they do not escape,” K Ajay Kumar, Assistant Commissioner of Police, Control Room, explained. The BCP will now be using Cobra teams (bike-borne quick reaction squads) as well to ensure the suspects are caught in time.
Talking about the project, Kumar said, “It is still a pilot project and we are working on improving some of the grey areas.” These facial recognition cameras are currently located at the Kempegowda Bus Stand, Yeswanthpur Railway Station, Orion Mall, Mantri Mall and Banashankari Bus Stand. Fifteen cameras are set up on three roads at the above mentioned locations. Each road has five cameras placed.
The number plate recognition cameras are located between the Central Business District and Palace Road, Yogeshwara Nagar to Manyata Tech Park, TC Palya to Hoskote, HSR layout to Silk Board, and Bhadrappa Layout to Hebbal. The CCTV surveillance company, which is working with the police, has also been able to use existing, ordinary CCTV cameras and modify them into all-purpose cameras, Kumar said. Currently, there are 603 PTZ (Pan Tilt Zoom) cameras and 128 bullet cameras across the city.
According to Gagan Jain, CEO and founder of Cyber Safe, a security company, the accuracy of facial recognition cameras depends on the algorithms fed into it. “If there are some criminals registered in the crime records, their three-angle photos are documented. This data needs to be fed into the system before-hand for the CCTV camera to recognise the person, while tracking them or if they have stolen a vehicle. Depending on how good the algorithm is, it can determine age, name, height and weight of the person,” Gagan says.
Speaking to City Express, Pradeep Kushwaha, head of Public Safety Business Unit at NEC, a Japanese ICT solutions company, which is working with BCP, said, “Face recognition is not just video analytics or image matching; rather, it is an advanced biometric technology. That is why many Artificial Intelligence (AI) based image matching face recognition solutions are able to perform in a demo environment, but fail to give results in real life scenarios.”
Explaining how their solution works, Kushwaha said, “It is used for real-time surveillance using CCTV video feeds, and also for post-incident investigation by searching for known watch-list data or missing-persons face data with a suspicious or missing person's face image. NEC’s face recognition solution performs under difficult conditions, like change in angle of face, lighting conditions, ageing, change in facial expression, hair style or beard, adding or removing glasses and other changes. The most important parameter is that its performance or accuracy does not get impacted much when the database grows to a significantly larger size.”