AI at the Edge Detection

Outdoor perimeter detection is one of the hardest jobs in automated security.

A perimeter camera has to detect real intruders at distance, across large outdoor areas, in darkness, weather, wind, vibration, low contrast, and changing thermal conditions. It has to do that early enough to support a response, accurately enough to avoid missed detections, and reliably enough to avoid the flood of false alarms that causes operators to stop trusting the system.

That level of performance does not come from treating a thermal camera like an ordinary video camera.

SightLogix was designed around a different principle: analyze the richest thermal data first, inside the camera, before that information is compressed, converted, delayed, or sent somewhere else.

The reason starts with the nature of thermal imaging itself.

A thermal camera does not start with a normal picture. It starts with raw temperature information from the scene. The grayscale image an operator sees is a display version of that data, created so people can interpret it visually.

SightLogix SightSensors analyze the raw thermal stream before that conversion happens. The camera uses the original thermal data for machine vision first, then creates a separate video stream for human viewing. The operator sees a useful surveillance image, while the camera uses the richer data needed for detection, classification, stabilization, and geospatial analysis.

That design choice has been foundational to SightSensor cameras from the beginning. It is what allows SightLogix systems to detect at long range, preserve alarm accuracy, reduce nuisance alarms, protect large perimeters with fewer devices, and generate alerts operators can count on.

Most surveillance video is optimized for people to look at. SightLogix is optimized first for the machine decision: is something moving, is it relevant, where is it, what is its size, how fast is it moving, and does it represent a real perimeter threat?

That sequence matters. The camera is not asking a downstream system to interpret a compressed surveillance image after critical information has already been discarded. SightLogix performs primary detection where the original thermal data still exists: inside the camera.

The result is faster, more dependable detection at the edge, especially when targets are small, distant, partially obscured, or difficult to see against a cluttered background.

Analyzing raw thermal data inside the camera requires more than a standard video processor. SightLogix cameras are built with the onboard processing needed to evaluate large volumes of thermal data in real time, at distance, and without relying on cloud processing.

• 120 Mb/sec of uncompressed thermal video analyzed on the camera
• 129 million thermal data points processed per second
• Image stabilization before analytics are applied
• Under 200 milliseconds for detection and classification
• No cloud dependency for primary intrusion detection

These capabilities help SightSensors generate trusted alerts quickly, locally, and reliably, without pushing primary detection to an offsite server or downstream video analytics platform.

For security-sensitive sites, primary intrusion detection cannot depend on a round trip to the cloud. SightLogix performs primary intrusion detection inside the camera, without relying on an offsite AI server, cloud analytics platform, or remote processing system.

That reduces latency, bandwidth demand, and exposure to network disruption. If a network connection is limited, interrupted, or intentionally cut, the camera still has the intelligence needed to detect and generate alarms locally.

For utilities, data centers, government facilities, logistics operations, airports, and other critical-site perimeters, that local intelligence supports faster response and greater confidence when action is required.

Edge detection is the first step in the SightLogix perimeter security workflow. When activity is detected, SightLogix analytics help determine whether the object is relevant. Geospatial intelligence calculates real-world location, size, speed, and direction. SightTracker PTZ can then move to the target location, zoom in, and provide visual assessment as the event unfolds.

The result is not just a detection. It is a faster, more useful alert that helps operators understand what is happening, where it is happening, and how to respond.

SightLogix edge detection is not just about where processing happens. It is about preserving the information needed to make reliable automated perimeter decisions.

By analyzing raw thermal data inside the camera before it is compressed or converted for display, SightLogix gives the system the information it needs to detect earlier, reduce missed detections, suppress nuisance alarms, calculate real-world target location, and support automated visual assessment.

That is how SightLogix delivers trusted alerts at distance across large outdoor perimeters — with fewer devices, lower infrastructure burden, and greater confidence when action is required.