When NuView they announced this I got flooded with questions from my customers and colleagues on whether this was going to work or not, and honestly I wasn't sure, but after seeing this it sounds like there could be something real here! I know it won't be able to get the point densities that we need for really high accuracy plotless forest inventories, but it's going to be a great source of additional data points for a large number of EO solutions. Excited to see how this evolves. Great interview Kevin, keep them coming!
Love Steve's thoughts and comments around increasing diversity in the industry, getting young people interested in forestry as a career as well as well as how climate change is evolving forestry management. The discussion on carbon credits was so thought-provoking and I think should be covered in the mainstream media more. Wonderful discussion, gentlemen!
Problem with ITC is properly segmenting the tree itself (segmentation problem), the paucity of training data (even 5000 field surveyed crowns is a tiny amount of training data compared to what machine learning thrives on) as well as the purity of the training data (badly identified and/or segmented crowns pollute the classifier during training) are the issues in our forests. Until that nut is cracked ITC will not be operational is the threshold is 90% (which we can consistently do for hardwood/softwood and ALS currently). Enjoyed this conversation as well.
