Image classification is the process of categorizing image pixels into specific classes. It is used across many fields such as facial recognition in social media apps and monitoring remote areas via drones. This process can be done on several image types such as hyperspectral, RGB, and grayscale images.
Hyperspectral images have many channels that span across the electromagnetic spectrum which allows them to convey more information than RGB images, which have only three channels. Hyperspectral imaging results in thorough spectral information which allows for distinguishing between materials. This is useful for military target detection or identifying mineral and wetland properties. Our goal is to apply several supervised learning algorithms to classify hyperspectral images and sets of RGB images. More specifically, support vector machines, random forest, convolutional neural networks, and k-nearest neighbors will be applied. We will then compare the accuracies and computation times of the various methods. Additionally, we will vary the hyperparameters in the algorithms to optimize the accuracies.