🤖 AI Summary
This work addresses land-use classification of onboard remote sensing imagery under resource-constrained, high-noise, and low-power in-orbit inference conditions. We systematically evaluate and optimize visual Transformer architectures for spaceborne deployment. For the first time, we comparatively assess ViT variants (EfficientViT-M2, MobileViTV2), CNNs, and ResNets in an Earth observation (EO) satellite environment, proposing a model selection paradigm centered on energy efficiency and noise robustness. Experiments on the UC Merced dataset show that EfficientViT-M2 achieves 98.76% classification accuracy with only 10 seconds inference time per image and 1000 seconds training time. Compared to MobileViTV2, it reduces power consumption by 63.93% and improves robustness score to 0.79—yielding the best overall performance. Our results establish EfficientViT-M2 as a new deployable, highly robust, and energy-efficient vision backbone benchmark for intelligent onboard remote sensing.
📝 Abstract
This study focuses on identifying the most effective pre-trained model for land use classification in onboard satellite processing, emphasizing achieving high accuracy, computational efficiency, and robustness against noisy data conditions commonly encountered during satellite-based inference. Through extensive experimentation, we compare the performance of traditional CNN-based, ResNet-based, and various pre-trained vision Transformer models. Our findings demonstrate that pre-trained Vision Transformer (ViT) models, particularly MobileViTV2 and EfficientViT-M2, outperform models trained from scratch in terms of accuracy and efficiency. These models achieve high performance with reduced computational requirements and exhibit greater resilience during inference under noisy conditions. While MobileViTV2 has excelled on clean validation data, EfficientViT-M2 has proved more robust when handling noise, making it the most suitable model for onboard satellite EO tasks. Our experimental results demonstrate that EfficientViT-M2 is the optimal choice for reliable and efficient RS-IC in satellite operations, achieving 98.76 % of accuracy, precision, and recall. Precisely, EfficientViT-M2 delivers the highest performance across all metrics, excels in training efficiency (1,000s) and inference time (10s), and demonstrates greater robustness (overall robustness score of 0.79). Consequently, EfficientViT-M2 consumes 63.93 % less power than MobileViTV2 (79.23 W) and 73.26 % less power than SwinTransformer (108.90 W). This highlights its significant advantage in energy efficiency.