Comparing ANN and Other Machine Learning Algorithms in the Study of Feature Classification Based on Multi-spectral Remote Sensing

Abstract

Due to the complexity and diversity of features, accurately recognizing their classification accuracy is significant for remote sensing data processing. Machine learning algorithms have high accuracy and can be applied to a variety of classification problems, providing excellent solutions to the problem of feature classification. In order to achieve fast and accurate classification of features and screen out the best model, this paper takes forests and airplanes as an example and compares and analyzes the accuracy effects of Support Vector Machine (SVM), Random Forest (RF), Artificial Neural Network (ANN) and AdaBoost (Adaptive Boosting) on feature classification based on multi-spectral remote sensing images collected by aerial. The results show that among the four models, SVM has the highest average accuracy, the best stability, and the fastest solution time, and can be used as the optimal model. ANN and AdaBoost are the next best, and RF is the worst. This paper can provide reference and guidance for feature classification model selection.