vb44/HMM-MOS — reverse-engineered prompt

Build me a small C++ command line tool for moving object segmentation in LiDAR point cloud sequences. I want it to take a folder of scans plus the sensor poses, build a voxel based local map, raycast each scan, estimate free and occupied space, then use a Hidden Markov Model filter to decide which points belong to moving objects.

Keep it learning free, so there’s no model training or neural network setup. Put the main settings in an easy config file, like voxel size, sensor range, uncertainty, probability threshold, and the local window length. The tool should save per point labels for static versus moving objects and print simple progress and timing information while it runs.

Please include a clear example for running it on common LiDAR MOS datasets like HeLiMOS, Sipailou Campus, or Apollo style data, and add a short README explaining the method, build steps, expected input format, output format, and how to tune the config. Look up current docs online if you need to.