Surveillance AI & Video Intelligence Glossary

A standard application-level network protocol designed to establish and control real-time media streams between IP cameras, NVRs, and AI video analytics software pipelines.

An open industry standard that defines standardized communication protocols (such as Profile S for streaming and Profile T for advanced video) between IP surveillance products.

A specialized computing appliance that captures, encodes, and records digital video streams transmitted over an IP local area network from IP security cameras.

A recording unit that receives analog video feeds via coaxial cables and digitizes them for storage, often integrated into modern AI platforms via video encoders or hybrid setups.

A hybrid digital video recorder capable of simultaneously ingesting analog, HD-over-coax (HDCVI, AHD, TVI), and digital IP camera streams on the same chassis.

Enterprise software that manages multi-camera recording, video playback, permission management, and camera health monitoring across distributed facilities.

Running deep learning and computer vision inference directly on localized hardware appliances within a facility's local network, eliminating external cloud bandwidth consumption.

The process of running live video frames through a pre-trained neural network to detect, classify, and track objects (such as people, vehicles, or safety violations) in real time.

An erroneous detection event triggered by non-threatening stimuli (such as headlights, windblown foliage, shadows, or animals) that AI spatial filtering is designed to suppress.

An optical character recognition (OCR) and computer vision pipeline that automatically reads and records vehicle registration plates from camera feeds at facility gates.

A software-defined boundary line configured on a camera stream that triggers an alert when a person or vehicle crosses in a specified direction, replacing mechanical tripwires.

Automated surveillance capabilities that monitor outer fence lines and boundary perimeters to detect unauthorized physical incursions before intruders reach facility buildings.

A spatial computer vision rule where a polygonal area on a camera stream is marked as restricted, triggering immediate alerts if unauthorized personnel or vehicles enter.

An imaginary or rendered rectangular coordinate frame enclosing a detected object (person, vehicle, or tool) within an image frame, indicating its spatial position and dimensions.

An automated camera health check that identifies when a camera lens is spray-painted, physically covered, displaced, or significantly defocused.

A physically and logically isolated computer network that has zero physical or wireless connections to the public internet, commonly mandated in defence and heavy industry.

The architectural principle of capturing, processing, and storing only the specific event metadata and short alert clips required for security operations, avoiding perpetual hoarding.

A security mechanism that restricts system access and stream viewing permissions based on an individual operator's verified organizational role and operational need.

The physical pixel distance between human eye centers on an optical camera sensor, used as a key technical threshold to determine facial verification feasibility.

India's statutory data protection law governing digital personal data, establishing fiduciary responsibilities and individual consent requirements for identifiable biometric video.

Hardware modules that convert software alert triggers into physical electrical signals, activating sirens, strobe lights, or automated boom barrier gates upon an AI event.

The frequency at which consecutive image frames are captured and streamed by a camera, typically balanced between 10 to 25 FPS to optimize AI inference throughput.

A networking technology (IEEE 802.3af/at) that delivers both electrical power and data connectivity to IP surveillance cameras over a single Category 5e/6 Ethernet cable.

A motorized surveillance camera capable of directional and optical zoom control, often paired with fixed wide-angle cameras for automated target tracking.

An advanced deep learning inference optimizer by NVIDIA that accelerates neural network models on edge GPU hardware to deliver sub-second video analytics latency.

Multi-layered neural network architectures (such as convolutional networks and vision transformers) trained to recognize complex spatial patterns and semantics in raw video frames.