I work with photogrammetry and lidar for an engineering firm. Aquisition, processing and extraction/preparation.
Quality ground classification is the most common, straightforward task that's applicable for me. Noise identification and removal, ground classification and addressing residual vegetation in the ground class. Actually, good classification all around makes EVERY following task much, much easier, so my main interests in lidar -always- revolve around improvements to point cloud classification.
Random ideas about little projects you could work on to get some experience:
* Improvements to ground classification, in different environments
* Simplification of EG (Existing ground) surfaces to make them more usable by end users, with minimal impact to precision (both raster and vector file formats)
* Vectorization of features (for instance, the way most software generates building footprint polygons from lidar point clouds is not acceptable for my industry, beyond the early stages of planning. A building polygon should have vertices at obvious angle points, not 5,000 vertices. I find this to be true for most automatic feature extraction tools, although the paint markings and curb extraction tools work pretty well with good lidar datasets).
* Anything to do with improvements to the classification of electrical transmission line features, identifying catenaries and clearances, etc.
* Change analysis (spatiotemporal changes between lidar datasets) and design compliance / design deviations (spatial deviations between digital design and real world).
* Tree inventory (tree counts, location, class (deciduous vs coniferous), diameter at breast height, etc). This is actually one I've been wanting to play around with more, using SLAM datasets. You could probably request relevant datasets from someone like GeoSLAM for this. If I was starting a new nerdfun side project, this is what I'd be playing with.
*Identification of beaver dams in topobathy lidar datasets.
*sediment transport visualization (Basically the change analysis I mentioned previously, but I'm always daydreaming about better tools to visualize and animate these types of deliverables
That's just a few off the top of my head. Also, it might be more fun for you to contact the equipment manufacturers and request sample datasets, instead of exclusively using downloadable datasets from something like USGS or DOGAMI. The higher point density (500 - 1200+ pp sq m) are more fun to play with. 😉
