Core Concepts
The 3MOS dataset is a large-scale multi-source, multi-resolution, and multi-scene dataset for optical-SAR image matching, containing 155K image pairs from 6 commercial satellites with resolutions ranging from 1.25m to 12.5m, and classified into 8 different scenes.
Abstract
The 3MOS dataset was constructed to encourage the design of more general multi-modal image matching methods. It consists of 155K optical-SAR image pairs, including SAR data from six commercial satellites (GF3, TerraSAR, ALOS, Radarsat, SEN1, and RCM) with resolutions ranging from 1.25m to 12.5m. The data has been classified into eight scenes: urban, rural, plains, hills, mountains, water, desert, and frozen earth.
The dataset construction process involved data collection and preprocessing, manual image registration, image cropping, and scene classification. Extensive experiments show that existing state-of-the-art methods do not achieve consistently superior performance across different sources, resolutions, and scenes. Additionally, the distribution of the data has a substantial impact on the matching capability of deep learning models, proposing the domain adaptation challenge in optical-SAR image matching.
The 3MOS dataset is expected to play a role in applications like multi-source satellite image fusion and visual navigation, where high-precision image matching is a crucial prerequisite.