End-to-End Big Data Platform with Video Intelligence for Smart Cities


Infocomm - Video/Image Analysis & Computer Vision
Infocomm - Smart Cities


Problem Statements:
With the development of smart cities, there are the following challenges:

  1. Insufficient big data analytics for video
  2. Video resources not being integrated into strategic planning/violation detection
  3. Ineffective application of video and image information
  4. Incomplete and untimely visualisation of a city's operations

Focused on the analysis of abnormal behavior through video, the solution is an industrial AI system with an independently-developed end-to-end video intelligent big data platform, as well as a deep learning algorithm development platform. Not only can it quickly update algorithm models iteratively, but it also enables the quick customisation and adaptation to various other computer vision use cases through transfer learning methods and a simple deployment experience.


The video AI system has 4 distinct and separate features:

  1. Urban Environmental Monitoring System
    • Monitors specific/industrial vehicles
    • No need to replace monitoring equipment. Able to use existing security and traffic cameras
    • Multiple AI models with accuracy rates of over 90%
    • Replaces on-site investigation by law enforcement personnel
    • Big data analyses information and predicts places where violations occur frequently, and detects the pattern of vehicle violations

  2. Traffic Violation Management System
    • Self-developed deep learning algorithm architecture to support all traffic violation scenarios
    • Cloud computing architecture, with clustered intelligent video analysis capabilities in a distributed architecture
    • A single server supports a maximum of 40 channels of 1080P real-time video analysis
    • Single channel videos are able to be used for more than 30 types of violation detection models at the same time
    • 1 day to complete rapid deployment city-wide, bypassing the need to spend several months in replacing the front-end equipment
    • Algorithm module integration, supports upgrade and expansion at any time
    • Cross-camera tracking and forensics to form a complete evidence chain, to realize the integration of forensics and prosecution
    • Big data time and space analysis together with in-depth exploration of the patterns of violation events help establish long-term governance

  3. Standard Management System
    • Machine identification + manual review.
    • 24-hour intelligent early warning, full-time video monitoring, 0 omissions
    • Existing hardware has high availability, lowering cost of implementation
    • Automated forensics of video pictures and long-term archiving to avoid disputes
    • Uniform machine evaluation criteria to avoid subjective judgment

  4. Internet Content Security System
    • 100+ types of violation analysis, combined with human review to ensure safe content
    • Innovative high-speed filter with coarse sieve + fine sieve process: able to handle tens of billions of tasks per day
    • Accuracy rate of 97.3% related to pornography, terrorism, and sensitive content, with 98% manpower savings
    • Realize 2-day iteration of model based on proprietary deep learning platform, followed up with the latest audit standards in a timely manner
    • Full coverage of common picture and video formats, universally applicable to various content scenarios


  1. Urban Environmental Monitoring System
    Application: Pollution control, specific vehicle identification, vehicle violation identification, vehicle flow behavior statistics

  2. Traffic Violation Management System
    Application: Identification of motor/non-motor vehicle violating at urban road traffic intersections

  3. Standard Management System
    Application: Suitable for video analytics of footage from factories and stores with multiple types of intelligent identification and customization of multiple product features

  4. Internet Content Security System
    Application: B2B online content review or network supervision by information and communication authorities

Potential Users:

  1. Transportation authorities
  2. Information and communication authorities
  3. Security providers
  4. Industrial operators
  5. Stores


The benefits that the video AI system provides as follows, compared to current competitors:

  1. Urban Environmental Monitoring System
    Better governance: increases efficiency of violation detection by more than 3 times
    Saves time: shortens implementation time by 90% using no additional equipment
    Cleaner air: helps reduce pollution and dust emitted by specific/industrial vehicles

  2. Traffic Violation Management System
    Safer streets: Accuracy rate of identification of traffic violation by motor vehicles exceeds 90%

  3. Standard Management System
    Save time and effort: reduces manual inspection work by 90%

  4. Internet Content Security System
    Safer internet at lower cost: 80% cost savings through automated identification of internet content
A Distributed Ledger Technology Application to Manage Food Supply Chain
Digital Hygiene Map System for Indoor Space
Remote Photoplethysmography for Contactless Vital Signs Monitoring
Advanced Analytics and Sensor Platform for Workspace Management
Operational Intelligence Platform For Optimization of Water Treatment
Energy Harvesting Smart Mat for Automation Control and Activity Monitoring
Automating Rheumatoid Arthritis Assessment via Deep Learning
AI-powered Intelligent Video Analytics Solution
Integrated Video Analytics Platform
Intelligent Personal Assistants And Software Platform For Online Information Processing