I have recently been working on a new approach to measuring spatially dense 3D landslide displacement using lidar DEMs. By using simulation powered self-supervised learning, we can train a site specific optical flow predictor capable of accurately computing the dense 2D horizontal displacement. The vertical component of displacement can then be computed via a remapping technique. Overall, this method (SlideSim) produces significantly improved accuracies compared to existing methods.
The full SlideSim journal paper can be found here.