This industrial-grade intelligent analysis device integrates video decoding, data transmission, snapshot storage, alarm video recording, and AI algorithms into a single unit. It can connect with standard network cameras to analyze video content in real time, accurately detect abnormal behaviors, and trigger alarm signals.
Supports 8-channel video stream target capture and comparison, with up to 100,000 targets, behaviors, and events for monitoring; AI-triggered snapshot events can link to audio-visual alarms.
Supports 8-channel behavior analysis, including human posture recognition and action data recording for behavior event detection.
Stores AI-recognized snapshots locally and records 5 seconds of video before and after the snapshot to a TF card.
Generates alarm images and short videos, with alarm information reporting support.
Supports video streams in H.264 and H.265 formats.
Compatible with standard ONVIF IP cameras and mainstream brands.
Supports RTSP direct video streaming and automatic real-time network connection monitoring.
Can recognize and capture 8 channels of network video simultaneously; each channel can run 2 different algorithms.
Algorithm recognition accuracy reaches 90%, which can be further improved through training; minimum detection resolution is 16×16 pixels.
Recognition distance ranges from approximately 3–8 meters, depending on the camera’s image size.
Algorithms use common deep learning frameworks, including DeepLearning, DL, Caffe, Theano, MXNet, Torch, PyTorch, and YOLOv5.
Supports connection to 8 third-party cameras compliant with ONVIF, RTSP, and GB28181 protocols.
Dual Gigabit Ethernet ports with dual network IP configuration support for various networking environments.
Optional 1×4G/5G module slot.
Supports web browser access and control via B/S architecture.
Lightweight, fanless design for flexible deployment; suitable for harsh environments with heavy dust, oil, and extreme temperatures.