No abstract

Hiroki

When a something occurs that could hinder a train passing a level crossing, obstruction warning signals are turned on to show that trains must stop. At present, this system relies on the driver's visual observation. This research aims to make this system more reliable and secure. This paper proposes a method to detect blinking obstruction warning signals and describes evaluation tests that were performed. Evaluation test results showed that these blinking signals could be detected with a detection rate of about 90 percent at a distance 600 meters under running conditions.

Recently, constructions of obstruction-warning signals and emergency buttons have been urgently needed at level crossings. However, the construction of these wayside facilities involves considerable inefficiency because it requires a joint survey or instruction to check these visibility from the train cab. Therefore, we have developed a simulator that overlays wayside facilities image on video sequences by estimating a three-dimensional position using our rail extraction method. In this paper, we present the concept of the simulator, details of the method, and an accuracy verification based on actual video sequences.

An obstruction-warning signal is an important piece of signalling equipment used at level crossings in Japan. However, it is not possible to check the visibility of such equipment during train operating hours, because this would require service disruption. This paper therefore proposes a method which employs near infrared LEDs, and can check the visibility of obstruction-warning signals during in-service hours. It has been proved that the developed image processing method, using a near infrared camera and a pan-tilt camera by feedback control installed on a vehicle, can correctly detect specific signal blinking. The paper then gives results from functional field tests.