Railway disaster monitoring system

An image processing unit of a railway disaster monitoring system for monitoring a foreign matter on a railway, the railway disaster monitoring system including a camera and the image processing unit that receives a railway image captured by the camera, the image processing unit comprising:

• a segmented image acquisition unit configured to obtain a plurality of segmented images including a rail from the railway image received from the camera and scale the segmented images to obtain segmented image blocks of a predetermined size;
• and
• a segmented image determination unit which includes a first deep neural network (DNN) discriminator trained with a deep learning neural network and inputs the segmented image blocks to the first DNN discriminator to determine whether the railway is in a normal state in which there is no foreign matter on the railway,
• wherein the segmented image determination unit further includes a second DNN discriminator that receives the segmented image blocks and further checks whether an abnormal state of the railway determined by the first DNN discriminator is correct when it is not determined by the first DNN discriminator that the railway is in the normal state, wherein the second DNN discriminator is trained with the deep learning neural network.

A rail management server of a railway disaster monitoring system for monitoring a foreign matter on a railway, the railway disaster monitoring system including

• a camera,
• an image processing unit including a segmented image acquisition unit that obtains a plurality of segmented images including a rail from a railway image received from the camera and scales the segmented images to obtain segmented image blocks of a predetermined size, and a first deep neural network (DNN) discriminator that determines whether the railway is in a normal state in which there is no foreign matter on the railway, and
• the rail management server connected to the image processing unit via a network,
• wherein the rail management server includes a second DNN discriminator trained with a deep learning neural network and configured to receive the segmented image blocks and further check whether an abnormal state of the railway determined by the first DNN discriminator is correct when it is not determined by the first DNN discriminator that the railway is in the normal state.

A rail management server of a railway disaster monitoring system for monitoring a foreign matter on a railway, the railway disaster monitoring system including a camera and a rail management server which is connected to the camera via a network to receive a railway image captured by the camera, the rail management server comprising:

• a segmented image acquisition unit configured to obtain a plurality of segmented images including a rail from the railway image received from the camera and scale the segmented images to obtain segmented image blocks of a predetermined size;
• and
• a segmented image determination unit which includes a first deep neural network (DNN) discriminator trained with a deep learning neural network and inputs the segmented image blocks to the first DNN discriminator to determine whether the railway is in a normal state in which there is no foreign matter on the railway,
• wherein the rail management server further comprises a segmented image area setting unit configured to set areas of the plurality of segmented images that are gradually reduced according to a perspective of the rail in the railway image received from the camera.

A rail management server of a railway disaster monitoring system for monitoring a foreign matter on a railway, the railway disaster monitoring system including a camera and a rail management server which is connected to the camera via a network to receive a railway image captured by the camera, the rail management server comprising:

• a segmented image acquisition unit configured to obtain a plurality of segmented images including a rail from the railway image received from the camera and scale the segmented images to obtain segmented image blocks of a predetermined size;
• and
• a segmented image determination unit which includes a first deep neural network (DNN) discriminator trained with a deep learning neural network and inputs the segmented image blocks to the first DNN discriminator to determine whether the railway is in a normal state in which there is no foreign matter on the railway,
• wherein the segmented image determination unit further includes a second DNN discriminator that receives the segmented image blocks and further checks whether an abnormal state of the railway determined by the first DNN discriminator is correct when it is not determined by the first DNN discriminator that the railway is in the normal state, wherein the second DNN discriminator is trained with the deep learning neural network.