The hybrid approach for railroad inspection has worked well and has reduced the continuous processing of large 2D laser data. This reduced the requirement of computational power, thereby also reducing the requirement of electrical power.
We believe that this solution can be further improved to run on a compact NVIDIA Jetson platform pre-loaded with trained models to cover additional surface defects. It can also be further enhanced to cover the track geometry and track component inspections.