Edge Algorithm Station

Edge Algorithm Station is an analysis device for back-end applications, equipped with a data stream AI chip, as well as AI algorithms and software, supporting 8-way video decoding and analysis, which can realize real-time AI analysis of pedestrians, vehicles, and the environment in surveillance video. Provide high-performance and cost-effective AI computing acceleration solutions for the chemical industry, construction sites, mines, grain depots, oil fields, and other scenarios.

Smoke Recognition
Facial Recognition
Flame Recognition
Intrusion Recognition
Fall Recognition

Smart phone APP notification
SMS/TTS Notification
Email Notification

RS232/RS485

12V DC

Need consultation?

Share your idea, We’ll find solution for you within 24 hours

ITEM	Specification
CPU Performance	≥2-core Cortex A72 @1.8GHz + 4-core Cortex A53 @1.5GHz.
Memory	4GB.
Storage	32GB eMMC + 1TB HDD.
AI Chip	• Domestic data-stream architecture chip.
	• Chip utilization rate (measured TOPS/peak TOPS) ≥60% (third-party test report required).
	• Peak AI computing power(INT8) ≥4TOPS.
Operating Temperature	• Operating: -20℃ to 60℃.
Operating Temperature	• Storage: -40℃ to 85℃.
Interfaces	RJ45 Ethernet port ×2, USB 3.0 ×2, USB 2.0 ×1, RS232 ×1, RS485 ×1, alarm input ×4, alarm output ×4, debug serial port ×1, power button ×1, reset button ×1, HDMI ×1, audio input ×1, audio output ×1, SD card slot ×1.
Protection Rating	IP40
Channel Capacity	• Single channel: Supports up to 10 parallel algorithms.
Channel Capacity	• 8 channels: Supports up to 16 parallel algorithms.

External Interfaces	Description
Network Interfaces	2 Gigabit Ethernet ports, adaptive to 10M/100M/1000M
USB Interfaces	2 USB 3.0 interfaces
Storage Interfaces	1 SATA interface (standard – equipped with 1T SSD), 1 M.2 M – Key interface
Display Interfaces	1 HDMI (only for temporary debugging, large – screen demonstration function is not supported)
Other Interfaces	1 SD card slot, 1 3.5mm coaxial audio output, 1 RS232, 1 RS485, 4 DIs, 4 DOs, 1 Debug Uart

Algorithm Function	Description
Smoking Detection	• Detects smoking behavior and triggers alarms; returns bounding boxes of detected heads.
	• Detects smoking individuals with head pixel size ≥60×600@1080p.
	• Configurable head detection thresholds, smoking classification thresholds, head quality thresholds, and side-face filtering thresholds.
Fence-Crossing Detection	• Detects fence-crossing behavior; returns human body bounding boxes.
Fence-Crossing Detection	• Detects individuals crossing fences with a minimum height of 100 pixels @1080p.
Crowd Gathering Detection	• Detects crowd gathering behavior; returns head bounding boxes.
	• Detects gatherings with individuals ≥120 pixels in height @1080p.
	• Configurable crowd size thresholds and head detection bounding box thresholds.
Fire Exit Obstruction Detection	• Detects obstruction of fire exits.
Fire Exit Obstruction Detection	• Configurable object filtering pixel thresholds and duration thresholds for objects remaining in the ROI.
Fighting Detection	• Detects physical fighting behavior.
Fighting Detection	• Configurable pedestrian detection thresholds, fighting classification thresholds, and alarm trigger frequency thresholds.
Fall Detection	• Detects falling incidents; returns body bounding boxes.
	• Detects falls for individuals ≥120 pixels in height @1080p.
	• Configurable pedestrian detection thresholds and height-to-width ratio thresholds.
Sleeping Detection	• Detects sleeping behavior (e.g., lying on desks).
	• Detects upper bodies ≥60×600 pixels @1080p.
	• Configurable upper-body detection thresholds and sleeping duration thresholds.
People Counting	• Counts personnel in designated areas; returns head bounding boxes.
	• Supports customizable alarm thresholds (e.g., over-limit, under-limit, or range-based alarms) and delayed alarms based on duration.
	• Detects heads ≥30×300 pixels @1080p.
Crowd Flow Statistics	• Tracks entry/exit changes in ROI; marks head bounding boxes for crowd flow.
	• Detects heads ≥50×600 pixels @1080p.
	• Configurable head detection thresholds and maximum dwell time thresholds.
Phone Usage Detection	• Detects phone usage (e.g., browsing, taking photos).
Phone Usage Detection	• Detects individuals ≥240×800 pixels @1080p.