Any minimum size objectives to detect objects of interest?
Cameron Stuart commented
I think I covered this in the examples included in the topic. There is no minimum detection criteria, but a high level of detail in the data would be beneficial. You might think of this new system as a method to augment (not likely to replace) the routine human track inspection activity. Can you detect and report conditions that are pertinent to the track inspection activity? If the system sees wide-spread deployment (1000s of units), we now have a partial track inspection with every train pass and, when datasets are combined, we have a high fidelity track change detection system. Situational awareness increases exponentially and, hopefully, so does the level of safety on the railroad system. The topic is really asking for a first technological step into an automated, high frequency safety inspection paradigm. Can we we leverage new and emerging technologies to reach a state where every locomotive adds to the body of safety knowledge (think BIG data). The industry has started on this path with discrete, autonomous measurement systems - Autonomous track geometry (ATGMS) and Vehicle/Track interaction (V/TI) systems. We (FRA) are also progressing rapidly with automated, AI-enabled track change detection technologies. All good, but the number of inspection passes is still limited. Leveraging the lead car of every train (locomotive) as a platform for a track safety "sensor suite" of some type is a logical way to increase the inspection frequency.