Investigating the role of UAVs and convolutional neural networks in the identification of invasive plant species in the Albany Thicket

The study aimed to determine whether plant species could be classified by using high resolution aerial imagery and a convolutional neural network (CNN). The full capabilities of a CNN were examined including testing whether the platform could be used for land cover and the evaluation of land change over time. A drone or unmanned aerial vehicle (UAV) was used to collect the aerial data of the study area, and 45 subplots were used for the image analysis. The CNN was coded and operated in RStudio, and digitised data from the input imagery were used as training and validation data by the programme to learn features. Four classifications were performed using various quantities of input data to access the performance of the neural network. In addition, tests were performed to understand whether the CNN could be used as a land cover and land change detection tool. Accuracy assessments were done on the results to test reliability and accuracy. The best-performing classification achieved an average user and producer accuracy of above 90%, while the overall accuracy was 93%, and the kappa coefficient score was 0.86. The CNN was also able to predict the land coverage area of Opuntia to be within 4% of the ground truthing data area. A change in land cover over time was detected by the programme after the manual clearing of the invasive plant had been undertaken. This research has determined that the use of a CNN in remote sensing is a very powerful tool for supervised image classifications and that it can be used for monitoring land cover by accurately estimating the spatial distribution of plant species and by monitoring the species' growth or decline over time. A CNN could also be used as a tool for landowners to prove that they are making efforts to clear invasive species from their land.Thesis (MSc) -- Faculty of Science, School of Environmental Sciences, 202