Minerals exploration is becoming more difficult, particularly because most mineral deposits at the surface of the earth have been found. While there may be a lot of sensing data, there is a shortage of expertise to interpret that data. This thesis aims to bring some of the recent advances in AI to the interpretation of sensing data. Our AI model learns one-dimensional features (lineaments) from two-dimensional data (in particular, magnetics surveys, maps of gravity and digital elevation maps), which surprisingly has not had a great deal of attention (whereas getting two-dimensional or zero-dimensional features is very common). We define a convolutional neural network to predict the probability that a lineament passes through each location on the map. Then, using these probabilities, cluster analysis, and regression models, we develop a post-processing method to predict lineaments. We train and evaluate our model on large real-world datasets in BC and Australia.
This repository contains all codes used in my Master Thesis. This program was developed under Python3, using Numpy, Keras, Tensorflow, Pillow, TKinter, Matplotlib and Scipy libraries.
We use 8 aerial images to train this model:
We designed and trained the following model using Keras and Tensorflow libraries. It starts from the input layer on the left which is consist of patches of size W × W × 8. Then we have a convolution 21 layer that creates a 3 × 3 convolution kernel that is convolved with the layer input to produce the output. A rectified linear unit (ReLU) is applied to the outputs of the convolutions. In order to reduce the dimensionality and to allow generalization in patches, we use 6 × 6 max pooling operations, which combine the outputs of neuron clusters at one layer into a single neuron in the next layer. We use a flatten layer that reshapes and merges previous hidden layers in the network into a single one-dimensional array. Finally, we use a fully connected neural network with two hidden layers with ReLU activations and one output layer of size one with Sigmoid activation.
We developed our own small GUI Applet to open datasets, train our model with different variables.
You can find more details in my thesis here.