Deep Learning Meets Satellite Images -- An Evaluation on Handcrafted and Learning-based Features for Multi-date Satellite Stereo Images

TL;DR

This study evaluates and compares handcrafted and learning-based feature matching methods on multi-date satellite stereo images, revealing traditional methods remain competitive while deep learning approaches excel in specific scenarios.

Contribution

It provides a comprehensive comparison of traditional and deep learning feature matching methods on satellite imagery, highlighting their relative strengths and limitations.

Findings

Traditional methods like SIFT are still competitive.

Deep learning methods perform very well in certain scenarios.

Learning-based methods show promising results for specific cases.

Abstract

A critical step in the digital surface models(DSM) generation is feature matching. Off-track (or multi-date) satellite stereo images, in particular, can challenge the performance of feature matching due to spectral distortions between images, long baseline, and wide intersection angles. Feature matching methods have evolved over the years from handcrafted methods (e.g., SIFT) to learning-based methods (e.g., SuperPoint and SuperGlue). In this paper, we compare the performance of different features, also known as feature extraction and matching methods, applied to satellite imagery. A wide range of stereo pairs(~500) covering two separate study sites are used. SIFT, as a widely used classic feature extraction and matching algorithm, is compared with seven deep-learning matching methods: SuperGlue, LightGlue, LoFTR, ASpanFormer, DKM, GIM-LightGlue, and GIM-DKM. Results demonstrate that traditional matching methods are still competitive in this age of deep learning, although for particular scenarios learning-based methods are very promising.