AI Video Surveillance Glossary

An AI capability that identifies objects left unattended in monitored areas. The camera learns the baseline scene and detects when new stationary objects appear and remain for longer than a configured threshold. Critical for transportation hubs, public spaces, and government facilities.

Also known as LPR (License Plate Recognition). An AI capability that reads vehicle license plates in real-time from video footage. Used for parking management, traffic enforcement, access control, and law enforcement. Advanced ANPR systems read plates at speeds exceeding 200 km/h.

ArcisAI's proprietary generative AI video search platform. ArcisGPT allows security operators to search surveillance footage using natural language queries instead of traditional timestamp and filter-based approaches. It processes queries against edge AI metadata for rapid forensic analysis.

AI analysis that detects specific human behaviors in video, including loitering, aggression, slip-and-fall, running, crowd formation, and other anomalous patterns. Behavioral analytics go beyond simple motion detection to understand what people are doing.

An edge-to-cloud gateway device that connects on-premises camera networks to cloud management platforms. The ArcisAI Bridge Device enables remote monitoring and cloud backup without requiring cameras to communicate directly with cloud servers, maintaining data sovereignty while adding cloud accessibility.

California Consumer Privacy Act / California Privacy Rights Act. State-level privacy regulations governing how businesses collect and use personal information, including biometric data from surveillance cameras. CCPA grants California residents rights over their personal data.

AI processing where video is transmitted from cameras to remote cloud servers for analysis. Cloud AI introduces network latency (typically 2-10 seconds), requires continuous internet connectivity, and incurs recurring per-camera subscription fees. Contrast with Edge AI which processes locally.

An AI capability that estimates the number and density distribution of people in open areas. Unlike people counting at fixed points, density estimation works across wide scenes. Used for event safety management, emergency planning, and public space monitoring.

The principle that data is subject to the laws and governance of the country or region where it is collected or stored. For surveillance, data sovereignty means keeping video footage within the deploying organization's physical and jurisdictional boundaries. Edge AI cameras inherently support data sovereignty.

A subset of machine learning using multi-layered neural networks to learn from large datasets. In video surveillance, deep learning models are trained on millions of images to recognize faces, license plates, objects, and behaviors with high accuracy.

Artificial intelligence processing that occurs directly on the camera device rather than on a remote server or cloud. Edge AI cameras contain dedicated neural processing units (NPUs) that run AI models locally, enabling sub-300ms alert delivery, offline operation, and data sovereignty since video never leaves the premises.

An AI technology that identifies or verifies individuals by analyzing facial features from video footage. Enterprise implementations use facial recognition for access control, VIP alerting, and security watchlist matching. Modern edge AI cameras process facial recognition locally for privacy compliance.

European Union regulation governing the collection and processing of personal data, including biometric data from surveillance cameras. GDPR requires lawful basis for processing, data minimization, and individual rights including data deletion. Edge AI processing simplifies GDPR compliance.

A technology that enables natural language queries across surveillance footage. Operators type descriptions like 'find all red vehicles near gate B yesterday afternoon' and the AI translates the query into structured metadata searches. ArcisAI's implementation is called ArcisGPT.

A visual analytics technique that aggregates movement data over time to show areas of high and low activity. In surveillance, heat maps reveal traffic patterns, dwell times, and underutilized spaces. Used in retail for store layout optimization and in facilities for space planning.

An architecture where cameras process AI analytics locally (edge) while using cloud services for optional functions like remote access, multi-site management, and long-term archival. Hybrid cloud ensures cameras operate at full capability without internet while providing cloud benefits when available.

A centralized operations facility that aggregates data from cameras, sensors, traffic systems, and emergency services for unified smart city management. ICCCs provide real-time situational awareness, incident management, and data-driven decision making for municipal operations.

The highest impact resistance rating on the IK scale, indicating a device can withstand 20 joules of impact energy (equivalent to a 5kg mass dropped from 40cm). Vandal-resistant cameras use IK10-rated housings for deployment in public and high-risk areas.

An AI analytic that detects unauthorized entry into defined zones. AI-based intrusion detection distinguishes between humans, vehicles, and animals, dramatically reducing false alarms compared to motion-based systems. Zones can be configured with different sensitivity levels and response rules.

Internet Protocol camera — a digital video camera that transmits data over an IP network. Unlike analog CCTV cameras that require dedicated coaxial cabling, IP cameras use standard Ethernet or WiFi connectivity. Modern IP cameras support resolutions up to 32MP and include on-board AI processing.

An Ingress Protection rating indicating a device is completely dust-tight (6) and protected against temporary submersion in water up to 1 meter for 30 minutes (7). Enterprise outdoor cameras require IP67 or higher for reliable operation in rain, dust, and adverse weather.

A specialized processor in cameras that converts raw sensor data into viewable images. Advanced ISPs perform noise reduction, wide dynamic range processing, color correction, and low-light enhancement. ISP quality directly affects AI analytics accuracy.

An AI analytic that detects when objects cross a virtual line drawn on the camera view. AI enhancement adds directional awareness (detecting crossing in one or both directions), object classification (alerting only on specific object types), and scheduling capabilities.

An AI analytic that identifies individuals or vehicles remaining in a defined area longer than a configured threshold. Used to detect potential surveillance, suspicious behavior, or policy violations in areas with expected transient traffic like ATM vestibules or loading docks.

See ANPR. License Plate Recognition is the alternate term for Automatic Number Plate Recognition, more commonly used in North America.

U.S. federal legislation. Section 889 specifically prohibits federal agencies and contractors from procuring video surveillance equipment from certain Chinese manufacturers including Hikvision, Dahua, and their subsidiaries. NDAA compliance has become a de facto enterprise standard beyond government.

North American Electric Reliability Corporation Critical Infrastructure Protection standards. A set of cybersecurity and physical security requirements for entities that operate the bulk electric system. Includes requirements for electronic access control, physical security of critical cyber assets, and incident reporting.

A specialized AI chip designed for running deep learning inference. In AI cameras, dedicated NPUs handle tasks like facial recognition, object detection, and behavioral analysis without using the main image processor. Typical enterprise camera NPUs deliver 2-16 TOPS (Tera Operations Per Second).

A dedicated hardware appliance that records video from IP cameras. NVRs provide centralized storage and management for camera deployments. Modern NVRs may include AI processing capabilities for cameras that lack edge AI. Contrast with DVR (Digital Video Recorder) used with analog cameras.

An AI capability that identifies and locates specific objects within a video frame, such as people, vehicles, packages, or weapons. Modern object detection runs in real-time on edge AI cameras, classifying objects by type and tracking their movement.

An open industry standard for IP-based physical security products. ONVIF profiles (S for streaming, T for advanced streaming, G for recording) ensure interoperability between cameras, VMS platforms, and other security devices from different manufacturers. Essential for avoiding vendor lock-in.

An AI analytic that accurately counts individuals entering and exiting defined zones. Bidirectional counting maintains real-time occupancy numbers. Used for building occupancy compliance, retail traffic analysis, and transport passenger flow. AI-based counting achieves 95-98% accuracy.

A technology that delivers both data and electrical power over a single Ethernet cable to IP cameras. PoE eliminates the need for separate power cables, simplifying installation. Standards include IEEE 802.3af (15.4W), 802.3at (30W), and 802.3bt (60-90W) for cameras with higher power AI processors.

Personal Protective Equipment detection — an AI analytic that verifies workers are wearing required safety gear including hard hats, safety vests, gloves, safety glasses, and respirators. Automated PPE compliance monitoring replaces inconsistent manual inspections.

Pan-Tilt-Zoom camera — a motorized camera that can rotate horizontally (pan), vertically (tilt), and optically zoom to provide flexible coverage. AI-enhanced PTZ cameras can automatically track detected objects, follow preset patrol routes, and zoom to areas of interest based on AI triggers.

The specific section of the NDAA that prohibits telecommunications and video surveillance equipment from Huawei, ZTE, Hytera, Hikvision, Dahua, and their subsidiaries. Applies to federal procurement and extends to contractors and subcontractors.

Visual AI analytics that detect smoke plumes and flame signatures in video footage. Works at greater distances than traditional smoke detectors, functions outdoors, and distinguishes between smoke, steam, and dust. Essential for large open spaces, outdoor areas, and industrial environments.

An Indian government certification body under the Ministry of Electronics and IT. STQC certification verifies that surveillance equipment meets Indian quality, security, and interoperability standards. Required for many government procurement projects in India.

An AI analytic that detects when unauthorized individuals follow authorized personnel through secured access points without presenting their own credentials. Essential for mantraps, server rooms, and restricted areas.

A measure of AI processing performance, indicating trillions of operations per second. Higher TOPS enables running more complex AI models simultaneously. Enterprise AI cameras typically range from 2-16 TOPS.

Software that manages video from multiple IP cameras, providing live monitoring, recording, playback, and analytics. Enterprise VMS platforms support thousands of cameras across multiple sites. Leading VMS platforms include Milestone XProtect, Genetec Security Center, and manufacturer-specific solutions.

A camera technology that handles scenes with both very bright and very dark areas simultaneously. WDR combines multiple exposures to produce balanced images. Essential for cameras facing windows, building entrances, or other high-contrast environments. Measured in decibels (dB); 120dB+ is considered true WDR.