IEEE Transactions on Geoscience and Remote Sensing · 2023 · 23 citations · 64 references
Natural SciencesEarth SciencesComputer VisionDigital ImagingRemote SensingChange DetectionDigital Image ProcessingLand Cover Change DetectionLand Use Change DetectionRemote Sensing Change DetectionSpatial Change DetectionChange Detection From Multispectral ImageryChange Detection From Remote SensingChange Detection From Synthetic Aperture Radar ImageryDeep Learning-based Change DetectionDomain AdaptationDomain Adaptation For Change DetectionSynthetic Aperture RadarSynthetic Aperture Radar Change Detection
Change detection on heterogeneous optical and synthetic aperture radar (SAR) images is soaring and plays a crucial role in monitoring land cover changes, such as disaster emergencies and natural resource monitoring. This is commonly recognized as a promising but challenging work due to the intrinsic differences in imaging mechanisms between the optical and SAR images. Recently, deep learning-based change detection methods based on two-step processing have attracted attention, i.e., first image translation between optical and SAR images to alleviate their modality differences and then change detection based on the translated images. However, image translation itself is a trouble task for the heterogeneous optical and SAR images. The unreliable image translation results further limit the accuracy of change detection. In this paper, to mitigate this problem, we propose a change detection model on domain adaptation by novelty integrating change detection and image reconstruction into a unified framework. Specifically, we first transform the optical and SAR images into an intermediate common domain for comparison. Moreover, cross reconstruction for optical and SAR images is designed to maintain the characteristics of the images and improve the performance of domain adaptation. In addition, a feedback guidance mechanism is circumspectly designed to co-optimize change detection and image reconstruction tasks. Extensive experiments were conducted on four publicly available datasets, the results demonstrate the effectiveness of our proposed method.
64
Unpaired Image-to-Image Translation Using Cycle-Consistent Adversarial Networks
Jun-Yan Zhu, Taesung Park, Phillip Isola et al. · 2017 · 21.9K citations · Full text
Generative adversarial networks
Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza et al. · Communications of the ACM · 2020 · 13.6K citations · Full text