Artificial Intelligence for Detecting People Inside Train Carriages

[Page created in 2024 for translation of original content]

In the series of articles labeled "R&D," we invite you to delve deeper into the research projects we're developing at Altametris. In this article, discover the research project around automatic detection of people inside train carriages using Artificial Intelligence.

The Challenge of Automatic Detection of People Inside Train Carriages

At the end of service, SNCF Transilien teams are required to inspect each carriage in the station before they are sent to the depot or parking area.

These inspections are lengthy and arduous for the agents performing them, and organizing them requires allocating numerous human resources, for added value to the operation that is still not well guaranteed.

Yet, these inspections are necessary to ensure that no conscious or unconscious passengers remain on board a train that is no longer in commercial service, with no passengers on board, or parked at the platform, thus inadvertently being taken to an industrial maintenance or service facility.

Altametris' Objective

SNCF Transilien and Altametris have partnered to develop and deploy an image analysis system capable of automatically detecting the presence of passengers in a Transilien train carriage.

Techniques, Technologies, and Methods Deployed

This system relies on artificial intelligence algorithms, specifically object detection algorithms.

The field of Computer Vision encompasses multiple techniques from various engineering and computer science fields. The different methods aim to replicate human vision to understand the content of images.

In the context of this solution, we have used object detection, which is a deep learning technique that detects the presence of an instance (recognizing an object) or a class of objects in a digital image or video stream.

Results and Development Outlook 

During this R&D phase, we have designed a system consisting of a server box directly connected to the train's data stream, which processes the stream on demand. Once a person is detected inside the carriage, they are tracked throughout the scan, and a notification of their presence and location is sent to the train drive. 

Did you find this article interesting? Do you want to learn more about this project? Do you need assistance in creating your industrial digital twin?