I have just started doing LiDAR point cloud research, and I feel that the processing algorithms for LiDAR point clouds are generally segmentation, reconstruction, completion, etc. This isn't very interesting. I have an idea: design a deep learning network so that the network can autonomously summarize formulas from point cloud data. For example, the potential theoretical formulas of leaf area index and tree diameter at breast height were summarized from forest point clouds. Is this possible? Are there any similar studies that we can learn from?
Compared with raster or grid data having the same pixel size, point clouds are somewhat complex and have irregular point geometries, therefore a more complex representation of 3D scenes. Performing deep learning and machine learning on such datasets is quite more challenging than regular gridded data. Point clouds first need to be classified into ground and non-ground points then a large number of structural attributes can be extracted. At this stage, I would recommend learning some publications on LiDAR point clouds and waveform (including my publications) so that you can establish a standard background on this topic before you can develop any robust research methods.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Graphs