With the accelerated pace of life, people struggling in social life have accumulated a lot of pressure. And some of them choose to commit suicide to escape this dilemma. According to the 2013 and 2014 Korea Transportation Safety Authority’s survey, ...
With the accelerated pace of life, people struggling in social life have accumulated a lot of pressure. And some of them choose to commit suicide to escape this dilemma. According to the 2013 and 2014 Korea Transportation Safety Authority’s survey, the public casualties caused by committing suicide on railways were about 54% and 50% of the public casualties. Therefore, an effective and inexpensive system for preventing suicide and crossing without permission accidents is needed.
We designed a viable and effective system for preventing accidents on railways by automatically detecting human object using deep learning algorithm. Railway images are captured using a camera installed on a drone and the detection of humans is done using a CNN algorithm. When humans are detected, an emergency braking request can be sent to the train control center in real-time. We choose a CNN model based on Depthwise Separable CNN algorithm for training and running. With this model named MobileNets presented by Google Inc, I tested the purposed method on 1123 railway images and got the recognition accuracy of more than 96% on 204 images, showing that the proposed method can be used as a good railway suicide prevention system.