ContactShield: how to monitor social distancing through Machine learning, GPS and BLE
In the turbulent era of COVID-19 it is important to keep pace with the up to date solutions and innovative Machine Learning platform in order to contribute to the better management for containing the risk of contagion through social distancing.
It is a challenge to keep the distance in the densely populated areas, but it is the ultimate way to prevent the spread of this disease.
Therefore, what can help in this situation is the tool for the detection of people’s distance in real time, scoped to notify the people as soon as possible when a social distancing rule is broken.
Platform our team is currently developing, Contact Shield, is dealing with this issue. Keep reading to check out the steps we undertook in order to have the final product ready.
First phase: BLE
During the first phase of the app development, we used Bluetooth LE technology, with the possibility for the user to cancel in any time access to the Bluetooth system. In this case, the app warned the user that the app system protection was missing.
It was in fact able to identify the devices that are excessively close to each other and, in the event of a positive report, it was able to signal to all users who have come into contact to carry out a preventive quarantine measures, limiting the need to prevent the closure of all the business activities.
Through the use of some devices capable of being able to communicate via Bluetooth LE technology, it was possible to identify limited local areas and activate various functions in the event of proximity, such as sending notifications to the user.
In order to provide a tracement model compatible with the Privacy By Design paradigm, Bluetooth LTE was used in order to register a contact between two devices in the given range.
The custom Bluetooth layer was responsible to both scan and transmit the ID of devices so, during a contact event, the ID of the other device, complete with time, date and approximate distance is recorded locally. All local contacts were then sent back to a central server regularly.
Second phase: Machine Learning
The initial phase proved to have the limits in the usage of the aforementioned technology. Therefore, we continued to use a different approach explained in the following paragraphs.
Because of that we are currently working on a Machine Learning approach by using computer vision. The idea is to detect people from a live video stream (it could be a security/surveillance camera) and to measure the distance between them, so in case they violate social distancing measures, they can be linked with others that have been in contact with.
One of the new features we developed is the possibility to pair with an external system for image recognition (and consequent reception of excessive proximity notification), like for example an internal surveillance system.
The user will reach a predetermined point in a building (which has an active system for image recognition that we call “The Gate”). The system recognizes the presence of a person, remembering the physical features — necessary for the re-identification throughout the structure during the day — and assigning an internal ID to her/him.
During a Contact Event, the system measures the approximate distance between the “bounding boxes” and produces a push notification alerting the users of excessive proximity to one another.
The system assigns each person unique IDs, by detecting a person based on their appearance (clothes, height, gait patterns) without collecting any personal information, for example, there’s no face recognition used.
The solution explained step-by-step
- The camera intercepts a “person”; the system assigns a unique ID;
- Through the mobile app, the person frames the QR code in order to connect their phone to the unique ID just generated;
- If the social distancing rules are not followed, users receive a notification on their phone.
Third phase: Indoor GPS
Through the partnership with one of the leading companies on the market, the evolution provides for the possibility of eliminating Bluetooth technology to be replaced by GPS, but with antennas installed Indoor, to allow optimal and precise monitoring.
If you want to know more about our approaches and/or you would like to suggest ideas or know more about privacy concerns and other topics, continue to follow us to stay up to day or contact us through our social media channels on Linkedin, Facebook or Instagram. We will be glad to reply and hear back from you.
