Face recognition entrance guard of campus gate
author:admin
time:2020-08-21
hits:2470
Face recognition all-in-one machine solution, based on high-performance hardware configuration, adapts to a variety of face recognition algorithms, has strong face recognition performance, supports a variety of recognition modes, efficiently responds to detection speed, and is equipped with complete background management software, which can be directly applied to various projects such as face brushing payment, access control, attendance checking, etc., and reduces the time and cost of research and production.
Provide a complete set of offline face technology services, including face detection, multi person tracking, face recognition, key point positioning, RGB live detection, binocular infrared live detection, occlusion detection, face quality judgment, gender and age judgment, etc.
The scheme builds face X1 face recognition machine + face recognition algorithm + cloud background management system
Recognition accuracy (30000 face database): 99.7%
Recognition accuracy (700000 face database): 95%
Threshold: 0.70 ~ 0.90
Detection time: 4ms ~ 20ms
Feature extraction time: 10ms-200ms
RGB silent biopsy time: 50ms-100ms
Binocular infrared biopsy time: 50ms-100ms
1: N comparison time: 1 million face database 200ms
3D face pose estimation time: 10ms-20ms
High performance hardware configuration for adaptive core
It adopts face-rk3399 high-performance motherboard, 8-inch full fit IPS screen, 800x1280 resolution, supporting multi-point touch; built-in wide dynamic binocular camera, 2 million pixels (1920 × 1080), more than 100dB wide dynamic range, 0.1Lux at f1.8 ultra-low illumination; all aluminum alloy high-precision CNC shell, no fan design, high-efficiency heat conduction
Efficient response detection speed
It adapts to a variety of face recognition algorithms smoothly and efficiently, and the recognition speed is less than 200ms; the recognition accuracy rate is higher than 99.77%.
Cortex-a8 / A9 / a5x / a7x series
Rk3288 / 3399 series
Intel x64 series
Hi3559 series
Nanopi Neo / M4 series
Raspberry PI 3 / 4 series
MTK series
High pass snapdragon series
Support multiple recognition modes
It supports up to 100000 face database for rapid recognition, and supports 1:1, 1: n, M: n recognition patterns, which can easily realize simultaneous recognition and detection of multiple people.
camera
Resolution: > = 300000
Video coding: YUV, nv21 / 12, mjpg, rgb888
Adaptive camera: USB wide dynamic camera, Mipi camera, infrared binocular high performance hardware configuration