81-83 Campbell Street, Surry Hills NSW 2010
Monday to Friday: 7AM - 7PM AEST
Mining companies are increasingly competing for a diminishing number of high-quality deposits. Mineable ore bodies tend to be concentrated in small areas, called mineral districts, and remote from large industrial centres. Prospective sites must be identified by remote sensing, geophysical prospecting methods like ground or airborne geophysics surveys and mapping, or field exploration using drilling and sampling techniques. Computer vision is a useful tool to assist visual inspection or digital photogrammetry when mapping is not practical (e.g., high cost). This article presents an overview of several computer vision algorithms that can help identify mineral deposits through data augmentation in mineral prospectivity prediction (MAP). In addition to the algorithm itself, this article also provides information on data preprocessing, training and testing methods.
Data augmentation in mineral prospectivity prediction is the process of using different computer vision algorithms to enhance existing data and/or adding new data to improve the accuracy of a computer vision system. This can be done by adding more samples, increasing the breath or scene coverage, or enhancing ground truth information on image features such as texture, colour, shape, and composition.
Data augmentation is a method of data enhancement, in which additional information, such as 3D point clouds and images, are used to improve the quality or value of your dataset.
The process of data augmentation can help increase spatial resolution by using the depth information acquired from stereo cameras. It can also be used to add texture or colour information acquired from RGB cameras and in-situ sensors. The use of data augmentation can be especially useful when it comes to extracting minute mineral deposits that are often invisible to the naked eye. With this technique, prospectors can identify a deposit without spending any time on site and without having to fill out lengthy forms requesting specific details about their find.
Mineral prospectivity prediction is a process that consists of a variety of data and information sources, including remote sensing, geophysical prospecting methods like ground or airborne geophysics surveys and mapping, and field exploration using drilling and sampling techniques. As mining companies continue to compete for fewer high-quality deposits, computer vision algorithms offer an important technology to help visually inspect or digital photogrammetry when mapping is not practical (e.g., high cost).
Computer vision algorithms can help identify mineral deposits through data augmentation in mineral prospectivity prediction (MAP) as remote sensing images are preprocessed. This article presents some of the most common computer vision algorithms that can help identify mineral deposits through data augmentation in mineral prospectivity prediction (MAP), including image segmentation, circular shape detection, and edge detection. In addition to the algorithm itself, this article also provides information on data preprocessing, training and testing methods.
The kriging algorithm is an iterative method that uses a weighted average to produce a prediction of the value at some point in space. In other words, it interpolates values using data points and their surrounding areas. It is used mainly for geoscience applications and is known as one of the most accurate methods of prediction.
A classification and regression tree (CART) is an algorithm that uses the idea of a decision tree to classify an input into one of multiple binary classifications. This article discusses data preprocessing, training, and testing methods for a CART and presents a brief overview of the CART algorithm.
The importance of digital marketing:
– 89% successful
– 80% more brand awareness
– Target your audience more precisely</p>
Variance reduction in MAP prediction. The training process of Random Forest (RF) is often hindered by overfitting, while the testing process can be biased. The article presents a classification model with RF-variance reduction (RFVAR), which enables more accurate and less bias assessment. This can be achieved by adjusting the voting factors of RF to reduce their influence on the decision function and to increase their influence on the error function.
Convolutional neural networks are a popular model for generating predictive models, but traditional convolutional neural networks require more data and computation than is available in most mining environments. The authors of this paper introduce object-based convolutional neural networks (OBCNN) which parallelize across the entire image matrix by using objects that may be found in images. These objects can be arbitrary shapes, such as circles or rectangles, and are not limited to shapes that strictly fall within the spatial dimensions of an input image. This processing method has been shown to work well on non-planar surfaces with various textures and depth effects.
Data augmentation is one of the most effective methods for improving the accuracy of mineral prospectivity prediction. It is a method that uses computer vision algorithms to improve the process of mineral prospectivity prediction by using an appropriate data set with a higher accuracy level.