Hello! We’re using a 2D orthomosaic generated from drone imagery to then port into Google Earth Engine to run several forms of supervised classification on the imagery. The orthomosaic is of a small island. We want to assess the performance of this classifier when applied to the imagery, but the outer edges of our orthomosaic product are predictably messier/less certain because of the reduction in overlap and lower number of tie points (especially since the outer edges are in open water). However, the performance of this classifier is of particular interest in open water, as we’re looking into the presence and absence of submerged species of vegetation.
Is there any way to extract a raster/map layer of the relative number of tie points at or near each pixel in the orthomosaic in order to perhaps create a filter that can remove areas of the orthomosaic with lower levels of tie points, preventing the classifier from being run on these lower-confidence, lower-overlap areas? Thanks all!