Say you needed to know the average height of every tree in a given forest. how would you go about getting that information without sending workers out to physically measure the trees?
Currently, the best computerized methods for measuring trees from photos are inaccurate and organizations must often resort to manual measurement. This is problematic as the height of trees in a forest is a very common datapoint in ecosystem studies, industrial planning, and many other use cases. Our team has worked with Professor Bogdan Strimbu from the College of Forestry to develop a method of identifying tree heights from overhead images using image processing algorithms and a neural network. First, our approach trains a neural network on existing tree height data paired with aerial imagery. Next, our approach finds the relative canopy size of the trees in our target image. Finally, we feed the tree heights into our trained neural network and label the target image with our estimated tree heights. Our method can provide tree height estimation at a fraction of the cost of traditional methods - a valuable breakthrough for the fields of forestry and environmental engineering.
