Manuscript received November 20, 2023; revised January 4, 2024; accepted February 19, 2024; published June 6, 2024

Abstract—Remote sensing technology and its applications have attracted the attention of researchers. Background variation and the small objects in remote sensing images make the classification process a challenging task. In several domains, Generalized Additive Models (GAMs) have demonstrated their ability to capture nonlinear interactions between explanatory variables and a response variable. This research evaluates the GAM with Scale Invariant Feature Transform (SIFT) for airplane remote sensing image classification. SIFT is a widely used local feature detection algorithm that performs best under scale and image rotations. We compared their performance with different methods, such as Harris-Stephens (HARRIS), Features from Accelerated Segment Test (FAST), Maximally Stable Extremal Regions (MSER), Oriented FAST Rotated BRIEF (ORB), and Binary Robust Invariant Scalable Keypoints (BRISK). To evaluate the results of the GAM with SIFT, Support Vector Machine (SVM), Discriminant Function Analysis (LDA), Quadratic Discriminant Function Analysis (QDA), and K-Nearest Neighbors (KNN) were applied. Accuracy rate, recall, precision, F-measure, and Receiver Operating Characteristic (ROC) curve were used as evaluation indexes. Based on the test dataset, SIFT features with the GAM increase precision, accuracy, recall, F-measure, and ROC curve compared to other applied classifiers. We show the performance of the applied airplane classification technique using two benchmark datasets from Google Earth, which are NWPU-RESISC-45 and UCAS-AOD.

Keywords—object classification, Generalized Additive Model (GAM), local feature descriptors, remote sensing images, Scale Invariant Feature Transform (SIFT)

Cite: Haidy S. Fouad and Hend A. Elsayed, "Evaluation of GAM Classifier Performance for Airplane Remote Sensing Images Based on SIFT Features," Journal of Image and Graphics, Vol. 12, No. 2, pp. 158-167, 2024.

Copyright © 2024 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC BY-NC-ND 4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.