COVID-19 Countermeasure

Video Analysis AI Solution


A video analysis AI solution that measures "closeness," "density," and "crowd size" in real time from camera images.

Social distancing and the degree of crowding and closeness can be analyzed in real time using network camera images and AI, making it an effective tool for preventing the spread of COVID-19.
The system accurately detects the number and location of people captured by network cameras in real time, and calculates the distance between people, the degree of crowding, and the flow of people in the crowd.

It provides a function to issue an alert when people get too close to each other, a function to count the number of people in a crowd to avoid close proximity, and a function to display the counted number of people in a graph.
The "Closeness Alert" function is suitable for use in offices and hospitals, while the "Crowd Flow Statistics" function is ideal for public facilities, commercial facilities, and event venues that require an accurate understanding of congestion based on the number and density of people. It can also be used as a technology that can provide safety and security when building smart cities.

Technologies used

This solution uses the following key technologies. All of them are our proprietary algorithms.

  • Pose Estimation Deep Learning
    (proprietary method, achieved state of the art results on public benchmarks, as of 2020/6. To be released later)
  • Special deep learning technology for person detection
    (highly accurate even with occlusion and overlap)
  • Person attribute classification deep learning
  • Person and object tracking technology
  • Fast and stable video streaming processing
  • Crowd counting deep learning

All of these elemental technologies are deep learning methods that can be used as stand-alone products, boasting industry-leading accuracy and speed.
By combining and tuning several of these technologies for COVID-19, we were able to develop the system in a short period of time.

We can also combine other proprietary component technologies such as "Person Detection and Re-Identification".
Customization is possible depending on the task.

  • 01

    Proximity alert

    Ridge-i's advanced person detection engine takes into account the posture of the person and estimates their exact position. It also detects when people are too close to each other.

  • 02

    Estimates the density of each area

    Displays the number of people per area in real time for each specified area. Detects congested areas.

  • 03

    Time-series display of the number of visitors

    Displays the number of visitors for a given period of time. It is possible to measure the effect of strict or relaxed quarantine rules.
    ※Real-time analysis is being conducted using street cameras in Hamamatsu City.

Overview An AI solution that detects the number and location of people on network cameras in realtime, and provides statistics on the distance between people, the degree of concentration, and the flow of people in a crowd.
Main Functions Time series traffic measurement, density estimation, close proximity alert, and attribute estimation (e.g., masked or unmasked, etc). Analysis results can be provided on YouTube.
Application environment Public facilities, office buildings, hospitals, commercial centers, event venues, transportation hubs, etc.
Other Customization is possible for one-time events, in-vehicle camera linkage, connection to existing cameras, etc.


Watch a demonstration of the video analysis AI solution on Ridge-i's Youtube channel

Closeness alerts

Real-time analysis of the distance between people using images from network cameras and AI, and alerting the user with a sound alert.

Identical person tracking

Identifies a person without prior registration. A unique ID is assigned to each person on the camera, and even if the person takes off his or her jacket or mask, it is still possible to determine that it is the same person.

Crowd Counting

An input/output demonstration of technology for estimating human density from existing camera images.

(1) Demonstration of "crowd counting" using our technology applied to images produced commercially by Adobe Stock.

(2) In this demonstration, YouTube's ZAZA Magazine Channel is broadcasting a live video feed, analyzing it in near real time, visualizing the degree of density using a heat map, and counting the number of people passing by during the live feed.

Graphical representation of acquired data (example)

The following graphs show the results of near real-time analysis of live streaming video from the ZAZA Magazine YouTube channel, and the counting of the number of people passing through. As an example of the count, the changes in traffic volume are tabulated hourly from the acquired data and displayed as a graph.

Real-time traffic volume on Kajiyama Street in Hamamatsu City (AI automatic counting)


Case study: A customer with a nationwide network of branches.

  • Before the introduction of our system:Delegated to each store manager the responsibility to take measures against the "three Cs" and Covid ( whether this is actually being done is unknown).
  • After the introduction of the system:Different levels of countermeasures are identified at each branch, and countermeasure methods are shared among branches to increase the quality of countermeasures (visualization and improvement).