Improvements to Feed-Forward Neural Networks Used to Classify Forest Type Covers Based on Cartographic Features

As computational power increases, deep neural networks provide an empirically effective way to classify large data sets based on a small labeled subset. Forest type covers are an example application of were these networks can be used. It is imperative for private and public land management agency’s to have accurate records of forest type covers for the regions under their management. However, due to legal and manpower limitations, this is not always possible. Neural networks provide a mechanism by which this data may be obtained. We present an improvement on feed-forward neural network architectures used to classify forest type covers based on cartographic features.