Convolutional Neural Networks (CNN) consist of various hyper-parameters which need to be specifed or can be altered when defning a deep learning architecture. There are numerous studies which have tested diferent types of networks (e.g. U-Net, DeepLabv3+) or created new architectures, benchmarked against well-known test datasets. However, there is a lack of real-world mapping applications demonstrating the efects of changing network hyper-parameters on model performance for land use and land cover (LULC) semantic segmentation. In this paper, we analysed the efects on training time and classifcation accuracy by altering parameters such as the number of initial convolutional flters, kernel size, network depth, kernel initialiser and activation functions, loss and loss optimiser functions, and learning rate. We achieved this using a well-known top performing architecture, the U-Net, in conjunction with LULC training data and two multispectral aerial images from North Queensland, Australia. A 2018 image was used to train and test CNN models with diferent parameters and a 2015 image was used for assessing the optimised parameters. We found more complex models with a larger number of flters and larger kernel size produce classifcations of higher accuracy but take longer to train. Using an accuracy-time ranking formula, we found using 56 initial flters with kernel size of 5×5 provide the best compromise between training time and accuracy. When fully training a model using these parameters and testing on the 2015 image, we achieved a kappa score of 0.84. This compares to the original U-Net parameters which achieved a kappa score of 0.73.

Data pre-processing for developing a generalised land use and land cover (LULC) deep learning model using earth observation data is important for the classification of a different date and/or sensor. However, it is unclear how to approach deep learning segmentation problems in earth observation data. In this paper, we trialled different methods of data preparation for Convolutional Neural Network (CNN) training and semantic segmentation of LULC features within aerial photography over the Wet Tropics and Atherton Tablelands, Queensland, Australia. This was conducted by trialling and ranking various training patch selection sampling strategies, patch and batch sizes, data augmentations and scaling and inference strategies. Our results showed: a stratified random sampling approach for producing training patches counteracted class imbalances" a smaller number of larger patches (small batch size) improves model accuracy" data augmentations and scaling are imperative in creating a generalised model able to accurately classify LULC features in imagery from a different date and sensor" and producing the output classification by averaging multiple grids of patches and three rotated versions of each patch produced a more accurate and aesthetic result. Combining the findings from the trials, we fully trained five models on the 2018 training image and applied the model to the 2015 test image. The output LULC classifications achieved an average kappa of 0.84, user accuracy of 0.81, and producer accuracy of 0.87. Future research using CNNs and earth observation data should implement the findings of this project to increase LULC model accuracy and transferability.

Abstract: As the global population rises, there is an ever-increasing demand for food, in terms of volume, quality and sustainable production. Protected Cropping Structures (PCS) provide controlled farming environments that support the optimum use of crop inputs for plant growth, faster production cycles, multiple growing seasons per annum and increased yield, while offering greater control of pests, disease and adverse weather. Globally, there has been a rapid increase in the adoption of PCS. However, there remains a concerning knowledge gap in the availability of accurate and up-to-date spatial information that defines the extent (location and area) of PCS. This data is fundamental for providing metrics that inform decision making around forward selling, labour, processing and infrastructure requirements, traceability, biosecurity and natural disaster preparedness and response. This project addresses this need, by developing a national map of PCS for Australia using remotely sensed imagery and deep learning analytics, ancillary data, field validation and industry engagement. The resulting map presents the location and extent of all commercial glasshouses, polyhouses, polytunnels, shadehouses and permanent nets with an area of >0.2 ha. The outcomes of the project revealed deep learning techniques can accurately map PCS with models achieving F-Scores > 0.9 and accelerate the mapping where suitable imagery is available. Location-based tools supported by web mapping applications were critical for the validation of PCS locations and for building industry awareness and engagement. The final national PCS map is publicly available through an online dashboard which summarises the area of PCS structures at a range of scales including state/territory, local government area and individual structure. The outcomes of this project have set a global standard on how this level of mapping can be achieved through a collaborative, multifaceted approach.