Drone based Mapping and Identification of Young Spruce Stand for Semiautonomous Cleaning

Abstract

We propose a novel method to locate spruces in a young stand with a low cost unmanned aerial vehicle. The method has three stages: 1) the forest area is mapped and a digital surface model and terrain models are generated, 2) the locations of trees are found from a canopy height model using local maximum and watershed algorithms, and 3) these locations are used in a convolution neural network architecture to detect young spruces. Our result for detecting young spruce trees among other vegetation using only color images from a single RGB camera were promising. The proposed method is able to achieve a detection accuracy of more than 91%. As low cost unmanned aerial vehicles with color cameras are versatile today, the proposed work is enabling low cost forest inventory for automating forest management.