AI-Powered Video Analytics

Similarity Search

Coming Soon. Quickly find people in video footage based on similarity in appearance.

AI video analytics extracts human appearance characteristics, such as the color and type of clothing, shoes, and headdress, and stores them in the database. You can select a person in the frame or upload a photo to quickly find similar individuals in the video footage from multiple cameras.

Privacy masking

Intelligent object masking in live and recorded video

Axxon One can automatically mask objects in live and recorded video, which is essential for compliance with GDPR and other data protection/privacy regulations. An intelligent, resource-optimized algorithm reliably masks objects in the camera’s FOV while using minimal computing power. You can also apply AI detection to hide only certain objects, e.g., humans.

AI Vehicle make, model, and color recognition

AI identifies and stores the make, model, speed, and color of vehicles in a traffic stream. You can use all of these criteria to find vehicles in recorded video from multiple cameras. This is essential when the license plate number of a vehicle of interest is unknown.

Behavior Analytics

Human posture recognition

Behavior Analytics recognizes situations of risk early on by detecting specific human postures: for example, a cashier’s raised arms, or a person crouching by an ATM. Based on Behavior Analytics, Axxon One helps detect violations of predefined social distancing restrictions as well. Swift alerts to first responders minimize the risks for individuals, groups, and facilities.

Object detection

High accuracy in complex scenes

A neural network accurately detects specific types of objects, e.g. humans or vehicles. This technology can filter out false alarms in complex scenes with a large amount of non-relevant details. You can apply any conventional video analytics (loitering, line crossing, object appearance and disappearance, etc.) to the detected objects.

The Neural Counter counts moving or static objects of a specific type within the scene, e.g. cars in a parking lot or people on the sales floor. This is a valuable tool for non-security-related applications.

Neural networks can meet the needs of a particular facility by learning from video material obtained onsite.

Smoke and fire detection

Detect fire early in open and large enclosed spaces

Intelligent fire and smoke video detection operates in areas where other types of sensors are ineffective, e.g. in open spaces. It provides early detection of fire hotspots which leads to a significant reduction in damage.

Facial recognition

Positive/negative list support

Configure automatic scenarios when a match is found. For example, the positive list may notify store personnel of VIP customer arrivals, while the negative list may indicate known offenders or shoplifters.

LPR / ANPR

Positive/negative list support

Configure automatic scenarios when a match is found. For instance, notify the operator if the recognized number is blacklisted. ANPR runs on the server side or on supported LPR/ANPR cameras.

Intel® Distribution of OpenVINO™ toolkit

Increase AI performance

OpenVINO™ is a toolkit for computer vision applications which extends workloads across Intel® hardware (including accelerators) and maximizes performance. Intel® Distribution of OpenVINO™ toolkit is applied for neural network inference in AxxonSoft AI analytics tools.

Axxon One supports the latest Intel® Vision Accelerator Design products with Intel® Movidius™ VPU:

• Mustang-V100-MX8 (HDDL)
• Neural Compute Stick 2

AxxonSoft benchmarked neural network inference performance using Mustang-V100-MX8: one HDDL card can handle 65 channels of Behavior Analytics.