Seibu Railway introduces a system that indicates a stop signal on a train approaching the railroad crossing when detecting abnormalities in the railroad crossing by AI or 3D image analysis by analyzing AI and 3D image analysis.Started.
Seibu Railway has developed a joint crossing system at a railroad crossing by analyzing a camera image that mainly detects "people" in a railroad crossing with multiple manufacturers for the purpose of improving the safety of railroad crossings.As a result of the accuracy verification on the field for about three years, the stable detection capacity has been confirmed, so we will start the introduction test at three relatively many people and bicycles.
In this test, we will mainly verify "link between anomalous detection system and a special signal emitting machine" and "Remote railroad crossing status."
検知画面（夜間）Special signal emissions (a stop signal on the train and inform the railroad crossing abnormality when the train is detected in the railroad crossing, mainly the abnormalities in the railroad crossing, mainly the train is approaching.Works) and informs the driver the danger.In parallel, the test of two types of devices, "Railroad crossing AI monitoring system" and "3D image analysis monitoring system", in parallel, in parallel.Furthermore, by grasping the status of the railroad crossing in real time by the camera, the function to support smooth safety confirmation is also verified.In addition, the monitoring system and the situation are grasped independently, and the special signal emitting machine operates even if communication of the situation is interrupted.
Until now, if "people" were left behind in the certificate of crossing, the holding of the emergency button by the person was sure to notify the train of abnormalities.This time, the railroad crossing AI monitoring system to be introduced and the 3D image analysis railroad crossing monitoring system has excellent detection performance of "people", and can be installed easier and cheaply than conventional disorder detection devices.
The Seibu Railway aims to introduce full -scale after 2022 based on the results of the introduction exam.
踏切滞留AI監視システムIn the railroad crossing AI monitoring system, the "person" in the railroad crossing is treated from the railroad crossing camera image and the shape of the object is recognized.With multiple AI algorithms: "object detection" that detects stays of objects such as cars and "skeletal detection" to detect people's movement and stagnation, automotive and people in the railroad crossing are quickly detected.In the skeletal detection, the skeleton of the person in the image is detected by the joint point extraction of the image by deep learning and the connection status of each joint point.By learning the strength of the connection between joint points, it is possible to detect people with high accuracy.In addition, a low -light camera is used for the AI camera, and image analysis can be performed at night.
Cooperative manufacturer: Oki Electric Industry, Marubeni Network Solutions Ikebukuro Line, Ikebukuro No. 9 railroad crossing (Toshima -ku, Tokyo) December 14, 2021 -March 31, 2022 Ikebukuro Line / Tokorozawa City (Tokorozawa City, Saitama Prefecture)) December 16, 2021 -March 31, 2022
3D画像解析踏切監視システムIn the 3D image analysis railroad crossing monitoring system, the image analysis with a 3D camera with a built -in left and right lenses can recognize the distance, height, and volume as well as human eyes by the parallax of the left and right cameras, and can recognize the distance, height, and volume within the railroad crossing.High -precision detection of the left behind is possible.Since there is no detection of light or shadow without volume, performances that are not affected by the natural environment can be expected.
Cooperative manufacturer: Computer system Research Institute Shinjuku Line No. 2 railroad crossing (Suginami -ku, Tokyo) 2021 -March 31, 2022
★ If you find a danger at a railroad crossing, press the emergency button of the railroad crossing disorder.