I'm a GIS Tech who frequently utilizes in house gather UAV photogrammetry for production/analysis. I've encountered this issue twice now with some rather large datasets. When using gathered UAV data of a project site, I'm often asked to create digital surface models (DSMs) similar to the elevation models that are available for public consumption through USGS. Understanding that data gathered from a standard digital camera are not LiDAR data, I found that processing it similarly results (most of the time) in a successful creation of a bald earth model. However, on at least two or three of my larger datasets, I've noticed that the algorithms used in the automated classification tool of the photogramtric generated point cloud act as if they've classified the entire dataset for ground points, only to find that it's classified a fraction of the dataset. I was wondering if anyone else was having this issue, and if so, did you find some explanation as to why it was happening and a fix?
I can provide further details upon request.