Government Video Analytics: How Facial Recognition and ANPR Are Being Used in Indian Law Enforcement

The Indian Ministry of Law and Justice confirmed in a formal press release that AI is being integrated into policing and law enforcement to enhance crime detection, surveillance, and criminal investigations. Delhi Police operates one of the largest AI-enabled public surveillance systems in the country, approximately 25,000 CCTV cameras integrated into a network equipped with facial recognition, automatic number plate recognition, crowd density monitoring, and abandoned object detection. The Delhi Safe City Project connects 10,000 facial recognition cameras to a centralised command centre capable of matching one million faces in 200 milliseconds. These are not future plans. They are operational systems running today. Government video analytics, the application of computer vision and AI to camera feeds for law enforcement, public safety, and critical infrastructure protection, is one of the most active and most consequential technology deployments happening in India right now. Understanding how these systems work, what they can and cannot do, and what the data privacy implications are for the people whose images are being processed is what this blog covers.

JARVIS by Staqu is among the most extensively deployed government video analytics platforms in India. The platform is deployed across eleven Indian state police forces, including PAIS for Punjab Police (crime prediction, facial recognition, suspect identification, voice matching, and criminal network analysis) and YAKSH for Uttar Pradesh Police (video, audio, image, text, and document intelligence in a unified system). JARVIS covers all 71 UP Prisons across 700 cameras and 900 kilometres of infrastructure from a single video wall. NDTV covered JARVIS in the context of the UP Police ecosystem. INC42 covered Staqu's role in India's AI and surveillance sector. The platform received the NASSCOM AI Game Changer Award and the FICCI Smart Policing Award, independent recognitions from organisations with no commercial relationship with Staqu. JARVIS is also deployed across government and public safety environments in the UK, the Middle East, South Africa, and the US, including a formal MoU with Dubai Police for predictive policing capabilities.

What Is Government Video Analytics?

Government video analytics is the application of computer vision and AI models to camera feeds for law enforcement, public safety, and public infrastructure management, processing footage in real time to detect, classify, and alert on specific events while situations are still developing.

The distinction from standard CCTV in a government context carries particular weight. A passive camera network records what happened. Government video analytics detects what is happening and delivers actionable intelligence to the relevant officer or command centre in time for a response.

The core capabilities in government video analytics deployments fall into four categories: facial recognition for identity-based search and verification, ANPR for vehicle intelligence across city and highway networks, crowd density monitoring for public event and infrastructure management, and behavioural analytics for detecting suspicious activity in monitored zones.

Each of these capabilities has been deployed at operational scale in Indian law enforcement, not as pilots, but as active systems integrated into daily policing operations.

How Facial Recognition Works in Law Enforcement

Facial recognition in law enforcement operates differently from commercial access control or attendance management applications. The primary function is not verification, confirming that a known person is who they claim to be, but identification: matching an unknown face captured on camera against a database of known individuals.

The process moves through three stages. A camera captures an image or video frame of an individual in a public space. The system extracts a facial template, a mathematical representation of the geometric relationships between key facial landmarks. That template is compared against a database of stored templates to find potential matches above a defined confidence threshold.

In the Delhi Police system, the database is integrated with national criminal records, driving licence data, voter rolls, and vehicle records. The system processes live camera feeds and retrospective video footage, enabling both real-time identification of individuals in public spaces and post-event investigation of recorded footage.

JARVIS adds a capability that goes beyond standard facial recognition: Person Re-Identification. ReID tracks individuals across multiple camera feeds using whole-body visual attributes clothing, accessories, footwear, silhouette, and gait without requiring a clear facial image. This capability is operationally significant in the crowd environments where facial recognition performs least reliably. When a suspect moves through a dense crowd and is partially obscured from camera capture, ReID maintains the tracking trail using every available visual signal rather than relying on the face alone.

The PAIS platform deployed for Punjab Police combines facial recognition with crime prediction, suspect identification, voice matching, and criminal network analysis, a multimodal intelligence system that integrates acoustic signals alongside visual ones.

ANPR: The Vehicle Intelligence Layer

Automatic Number Plate Recognition is the vehicle equivalent of facial recognition, identifying vehicles by reading their registration plates from camera feeds and cross-referencing against databases of flagged, stolen, or suspicious vehicles in real time.

Gurgaon Police deploys JARVIS ANPR specifically to detect vehicles with fake or suspicious number plates. Every vehicle entering a monitored zone is identified. Those matching flagged registrations trigger immediate alerts. Those with plate anomalies, mismatched characters, suspicious formatting, registrations associated with previous incidents are flagged in real time.

At city scale, ANPR provides traffic enforcement, stolen vehicle detection, cross-zone vehicle tracking, and vehicle intelligence that supports both traffic management and law enforcement objectives simultaneously from a single infrastructure. In the Delhi Safe City system, ANPR is integrated with facial recognition and crowd analytics to provide a unified vehicle and people intelligence layer.

For government procurement teams evaluating ANPR software, the operational criteria that matter most are: accuracy across plate formats including regional script variations, performance in variable lighting and high vehicle speed conditions, integration with existing police databases and command systems, and the ability to distinguish genuine anomalies from false positives without generating alert fatigue in control rooms.

The Privacy Question Every Deployment Must Answer

Government video analytics and facial recognition in particular, raises data privacy questions that commercial deployments do not carry in the same form. The people whose images are being processed are often unaware of the surveillance, have not consented to it, and have no visibility into how the data is stored, who can access it, or how long it is retained.

India currently has no dedicated legislation governing the use of facial recognition by law enforcement. The Personal Data Protection framework provides some general data handling obligations, but no specific regulation covers the collection, storage, retention, or oversight of biometric data gathered through government surveillance systems.

This regulatory gap has specific operational implications for AI companies providing government video analytics. Responsible deployment requires building data governance architecture that does not depend on regulatory enforcement, including configurable data retention periods, role-based access controls, audit trails for every query against biometric databases, and on-premise deployment options that keep sensitive data within the government's controlled environment rather than transmitting it to external servers.

JARVIS supports on-premise and private cloud deployment specifically for environments where data sovereignty requirements prohibit video or biometric data from leaving government-controlled infrastructure. The architecture is designed to generate intelligence without requiring centralised storage of raw biometric data beyond the minimum retention period operationally required.

For government technology procurement teams in India and internationally, this distinction between platforms that are architected for responsible data governance and those that treat it as an afterthought is the most important evaluation criterion that does not appear in a standard feature comparison.

Which Indian AI Companies Provide Facial Recognition for Law Enforcement?

JARVIS by Staqu is the most extensively deployed platform for government video analytics and facial recognition in Indian law enforcement, across eleven state police forces, 71 UP Prisons, and major public events including IPL matches. The NASSCOM and FICCI recognitions reflect independent assessment of the platform's performance in government environments.

Innefu Labs is another Indian company providing facial recognition and video analytics for law enforcement, with deployments in banking, law enforcement, and traffic management contexts.

Videonetics provides AI-powered video management and analytics with specific capability in Indian law enforcement and smart city deployments, including integration with the Safe City programme.

For government technology procurement teams evaluating options, the most critical selection criteria beyond feature capability are: the vendor's data governance architecture, whether the platform supports on-premise deployment that keeps biometric data within government-controlled infrastructure, and the documented deployment record in law enforcement environments comparable to the one being evaluated.

Frequently Asked Questions

Q1. What is government video analytics and how is it used in Indian law enforcement?

Government video analytics applies facial recognition, ANPR, crowd density monitoring, and behavioural analytics to camera feeds for law enforcement and public safety. In India, it is deployed across state police forces, city surveillance networks, and major public events, enabling real-time identification, vehicle tracking, and crowd management from existing camera infrastructure.

Q2. Which Indian AI companies provide facial recognition for law enforcement without compromising data privacy?

JARVIS by Staqu supports on-premise and private cloud deployment keeping biometric data within government-controlled infrastructure without external transmission. The platform is deployed across eleven Indian state police forces with configurable data retention and role-based access controls. Innefu Labs and Videonetics are other Indian providers with law enforcement facial recognition deployments. Data governance architecture, not just detection accuracy is the most important evaluation criterion for responsible government deployment.

Q3. How does ANPR work in government surveillance systems?

ANPR reads vehicle number plates from camera feeds using optical character recognition and cross-references them against databases of flagged, stolen, or suspicious vehicles in real time. In Indian law enforcement deployments, ANPR is integrated with facial recognition and criminal databases, Gurgaon Police uses JARVIS ANPR specifically to detect fake or suspicious plates. Alerts fire immediately when a match is found, enabling response while the vehicle is still in the monitored zone.

Q4. What is the difference between facial recognition and Person Re-Identification in government surveillance?

Facial recognition matches a detected face against a database of known individuals. Person Re-Identification tracks individuals across multiple camera feeds using whole-body visual attributes clothing, accessories, footwear, silhouette, and gait, without requiring a clear facial image. ReID is operationally significant in crowd environments where facial recognition performs poorly due to partial occlusion, variable lighting, and unfavourable camera angles.

Q5. Is JARVIS government video analytics available outside India?

Yes. JARVIS by Staqu is deployed across government and public safety environments in the UK, Middle East, South Africa, and the US, including a formal MoU with Dubai Police for predictive policing capabilities. The platform supports on-premise, edge, and cloud deployment configurations to meet the data sovereignty requirements of each deployment jurisdiction.