Demonstration of AI stagnation detection system at railroad crossings  NTT Communications|電経新聞

Demonstration of AI stagnation detection system at railroad crossings  NTT Communications

NTT Communications (NTT Com), Kanto Railway, Koshida Tech, and Yashimakizai are currently applying for a patent on the Ebihara railroad crossing on the Kanto Railway Joso Line (Moriya City, Ibaraki Prefecture, between Moriya Station and Shin-Moriya Station). Demonstration and verification for the prevention of railroad crossing accidents using the “AI retention detection system at railroad crossings” that detects stagnant objects regardless of size in real time with high accuracy has started.
This system uses a 5G network to transmit images from commercially available network cameras installed near railroad crossings to docomo MEC, and uses VAE and background subtraction technology to analyze them with AI. It is something that can be detected with high precision and in real time. If an object is detected, an alert will be sent to the facility manager.
Level crossing accidents account for 40% of railway operation accidents, with 217 incidents occurring in FY2021. Additionally, 42.9% of railroad crossing accidents are caused by pedestrians delaying crossing. To prevent railroad crossing accidents, it is common to install a level crossing obstruction detection device that detects stagnation on the tracks. However, while large objects such as automobiles can be detected with high accuracy, improving the detection accuracy of small objects such as bicycles, strollers, wheelchairs, handcarts, and their companions has been a challenge.