본문 바로가기 주메뉴 바로가기

(주)엠제이비전테크

Edge/Edge AI CCTV

Development background

A system that processes the image analysis algorithm, which was previously analyzed centered on the server, in a local control system

Features

  • 01

    Independent video analysis CCTV based on edge AI device

  • 02

    Used for purposes such as Parking Violation and Illegal Trash Dumping, local security, etc. through specialized object recognition functions

  • 03

    Securing existing CCTV compatibility and scalability in the form of camera integration or edge board module addition

  • 04

    Applicable to small-scale local governments or privately owned tourist destinations

Edge AI deviceIntelligent video analysis control system

  • NPU all-in-one
    • Real-time operations such as data generation, decision-making and action are processed instantly within the same hardware.
  • Small and light
  • Low price
    • Compatible with existing CCTV cameras for easy expansion
    • Competitiveness secured at a lower price compared to general PCs and devices dedicated to image analysis
  • High performance
    • Recognizes objects based on deep learning, and has an excellent object recognition rate compared to conventional motion-based object detection
    • Depending on the recognition object, it can be used for various purposes

Composition

  • Add-on module

  • Camera all in one

Expected effect of product introduction

  • Possible to build a system with a low budget
    • LPR(License plate recognition) system: Vehicle number identification camera + Edge (additional)
    • Accident prevention using cameras in restricted areas (pop-up, voice transmission)
  • Provision of specialized services
    • Using cameras installed in specific areas to utilize floating population and traffic (vitalization of commercial districts)
  • Securing various actual data
    • Securing various actual data by operating service server (cloud type) (personal information consent or shooting notice required)

Function

Edge AI device Equipped with deep learning-based image analysis algorithm

Edge AI device
CPU AmlogicA311D x4 A73 at 2.2GHz + x2 A53 at 1.8GHz
GPU ARM Mali G52 MP4 at 800MHz
NPU AmlogicA311D : 5.0 TOPS
INT8 inference up to 1536 MAC
RAM 4GB
eMMC 32GB
Wi-Fi 2T2R 802.11 ac with RSDB
Bluetooth V5.0
Test Condition Darknet53 / Yolo v3 / COCO Dataset / 8MP HDR Camera
FPS for Detection 7~8 FPS
  • Possession of deep learning-based object recognition technology

    Possession of deep learning-based object recognition technology capable of real-time processing

  • Edge AI device development completed

    Completed development of edge AI device with 5 TOPS (Trillion Operation Per Second) performance based on NPU capable of deep learning network processing (e.g., Apple's next-generation A14 chip is 11 TOPS)

  • Edge AI device-based smart home cam/Development of smart traffic information analysis device

    Edge AI device-based smart home cam

Construction status

CCTV Camera System with Edge AI Computing

  • Daejeon Wongol Crossroads Intersection Control

Certifications

  • MJVT 엣지 CCTV 1.0

    MJVT Edge CCTV 1.0

상단으로