Monitoring and modelling vulnerable wildlife populations using remotely piloted aircraft systems and machine learning

In this thesis a new method for monitoring wildlife using drones and machine learning was developed and used to accurately detect threatened koalas in natural, complex forest habitats for the first time. The new automated detection method was robust, efficient, and was not subject to biases affecting ground surveys or manual analysis of thermal images from drones. A new statistical modelling approach was also developed that allowed accurate estimates of abundance to be made from automated detections of wildlife in drone surveys