Comparison based on Multiple Machine Learning Algorithms in the Case of Image Classification

Abstract

With the upsurge of usage of the internet, the society has turned into a Big Data Era. Living in the world overwhelmed with data, the image is among the most common forms. Utilizing these images smartly is the big topic. Image Classification refers to the process of extracting information classes from a raster image that consists of multiple bands by analyzing the raster image. It is the basis for lots of applications, like radar and autopilot. Among tremendous channels for dealing with image, deep learning, Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN), is always the first thing flashing in peoples’ mind. In this paper, studies will be implemented and discuss the way to perform image classification utilizing machine learning algorithms. Four algorithms will be implemented, namely Random Forest, KNN, Decision Tree, and Naive Bayes. And at the end of the paper, this result shows that Random Forest Classifier has the best performance compared with the rest of the three algorisms. But it is still far away from the requirement for daily application use like radar. Poor performance and the long operation time make machine learning algorisms out of date when doing image classification. By this comparison, deep learning replaced machine learning and becomes the first choice do this task.