Rail Surface Inspection
Critical Asset Detection Algorithms
The following algorithms were developed for a major national railroad to automatically detect numerous fixtures and hardware components that are considered critical assets by the federal rail administration. For this approach, a video camera and GPS/IMU system were located on the back of a locomotive to passively capture the track and surrounding environment while the locomotive was in motion.
The following asset detection algorithms were developed by Anthony to support this project.
Turnouts are the locations where the railroad splits from a single set of tracks to multiple. Turnout detection automatically detects and catalogs the locations of turnouts.
Graded Crossing Detection
Graded crossings are any locations where a road, sidewalk, or pathway crosses the track surface. These crossings can range from narrow footpaths, to large highway crossings and have vastly different appearances.
Control Signal Detection
Control signals can indicate the direction, clearance, and allowed speed of a train through a junction. These can have a large variety of appearances including dwarf, pole-mount, cantilevered, and bridge mounted configurations.
Clearance Point Detection
Clearance points indicate the closest point to a turnout that two standard-size locomotives can pass by each other without colliding. These can be indicated by rail side indicators as wells as by hardware derailer systems.
Hardware derailers are located along track surfaces to intentionally derail a train in the event of unauthorized train movement or rolling stock. These devices come in a large number of configurations are are often indicated by track-side signage.
The background machine vision processing engine used to process this train data distributes the workload across numerous processing cores to provide large scale processing capability.