Particularly in light of unusually warm weather 1, 2, 3, the risks of railway accidents has dramatically increased over the last years. Inspections using special trains are costly and infrequent (twice a week, 4 ), while changes can happen over the course of a single day.
By mounting inexpensive portable imaging devices on each train, we can collect real-time image information on each of the tracks being crossed by processing the images using our Spark Image Layer.
The first question is how the data can be processed. The basic work is done by a simple workflow on top of our Spark Image Layer. This abstracts away the complexities of cloud computing and distributed analysis. You focus only on the core task of image processing.
Beyond a single train, our system scales linearly to multiple trains and computers to keep the computation real-time.
With cloud-integration and Big Data-based frameworks, even handling an entire train network with 100s of trains running continuously is an easy task without worrying about networks, topology, or fault-tolerance. Below is an example for 30 trains where the tasks are seamlessly, evenly divided among 50 different machines.
The images which fly past the train at hundreds of meters per second are rich in information about the tracks, structure, and even details to potential upcoming dangers. The first basic task is the segmentation of the tracks which can provide information on their separation, surface smoothness, and direction.
The segmented image above can be transformed into quantitative metrics at each time point. These metrics can then be processed to extract relevant quality assessment information for the tracks.