Improving Remote Sensing Scene Classification with Data Augmentation Techniques to Mitigate Class Imbalance

Ping Wang¹Xin Zhao¹Yuanhui Chen²Lili Zhan^2*

• ¹Qingdao Huanghai University, Qingdao, China
• ²Shandong University of Science and Technology, Qingdao, Shandong Province, China

The high-resolution remote sensing scene images can provide abundant information about ground objects. However, classical methods often fail to achieve satisfactory results for complex urban scene classification. Deep learning-based remote sensing image scene classification (RSSC) represents an important approach for understanding semantic information. Traditional methods are unable to meet the requirements of high-accuracy RSSC and are challenged by issues such as limited labeled samples and class imbalance, which may lead to classification bias in classifiers. This paper explores the feasibility of mitigating classification bias by reducing the imbalance ratio (IR) of the dataset. First, a class-imbalanced dataset was constructed using Very High Resolution (VHR) images, labeled into nine land use/land cover (LULC) categories. Second, comprehensive data augment methods (mirroring, rotation, cropping, HSV perturbation, and gamma transformation) were applied, successfully reducing the dataset's IR from 9.38 to 1.28. Subsequently, four architectures, MobileNet-v2, ResNet101, ResNeXt101_32×32d, and Transformer, were trained and evaluated on both class-balanced and classimbalanced datasets. The results show that the classification bias caused by class-imbalance was alleviated, with overall classifier performance substantially improved. Specifically, for the most severely underrepresented category (intersections), Precision and Recall improvements reached up to 128% and 102%, respectively, respectively, narrowing the gap with other categories and reducing classification bias. Furthermore, the average Kappa and overall accuracy (OA) increased by 11.84% and 12.97%, respectively, with reduced standard deviations in recall and precision demonstrating enhanced model stability. predefined semantic classes based on scene-specific information, representing high-level abstractions of scene content. However, traditional classification methods are limited by their reliance on low-level feature analysis, which restricts their capacity to extract high-level semantic information, thereby often failing to meet the accuracy demands of RSSC.Currently, the proliferation of deep learning has spurred numerous methodologies for remote sensing scene image classification, which can be broadly categorized into three types: autoencoder-based, Convolutional Neural Network (CNN)-based, and Generative Adversarial Network (GAN)-based approaches (