Summary
The rapid advancement of computer vision is changing the way many industries approach image analysis. Object detection, image segmentation, textual characterisation and other related processes are helping people extract more meaningful information and relationships from their new and existing imagery. Although more typically applied to photography, this paper shows how computer vision techniques can be meaningfully applied to geophysical and other geoscientific datasets.
An introduction to neural networks is followed by an application of computer vision via the ResNet-50 convolutional neural network to the GGMplus regional gravity model of Australia. This results in the quantitative characterisation of geophysical texture and the ability to spatially query areas for morphological similarity.