4 May 2020 in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Deep convolutional neural network (DeCNN) is considered one of promising techniques for classifying the high spatial resolution remote sensing (HSRRS) scenes, due to its powerful feature extraction capabilities. It is well-known that huge high quality labeled datasets are required for achieving the better classification performances and preventing over-fitting, during the training DeCNN model process. However, the lack of high quality datasets limits the applications of DeCNN}. In order to solve this problem, in this paper, we propose a HSRRS image scene classification method using transfer learning and DeCNN (TL-DeCNN) model in few shot HSRRS scene samples. Specifically, three typical DeCNNs of VGG19, ResNet50 and InceptionV3, trained on the ImageNet2015, the weights of their convolutional layer for that of the TL-DeCNN are transferred, respectively. Then, TL-DeCNN just needs to fine-tune its classification module on the few shot HSRRS scene samples in a few epochs. Experimental results indicate that our proposed TL-DeCNN method provides absolute dominance results without over-fitting, when compared with the VGG19, ResNet50 and InceptionV3, directly trained on the few shot samples.
27 July 2022 in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Synthetic aperture radar (SAR) satellite has been widely applied in real-time flood monitoring as that they are not affected by extreme weather conditions. However, there is no automatic method to quickly and accurately extract flood areas with SAR satellite images. In this article, a UNet combined with the attention mechanism (UNet-CBAM) method has been proposed for extracting flood submerged areas, and both Longgan Lake and Dahuchi in Poyang Lake Basin are selected as the test sites. Based on the polarization characteristics of two Sentinel-1A data of Poyang Lake, both UNet and UNet-CBAM extraction methods are utilized to extract the flood areas, respectively. Compared with traditional SAR image water extraction methods, simulation results demonstrate that UNet can obtain more satisfactory results in recall, precision, and ${F}_1$ -parameter, but it has no capability to guarantee continuity in edges and small bodies of water. Moreover, our proposed UNet-CBAM method can further improve recall, precision, and ${F}_1$ -parameter, respectively. Specifically, when compared with UNet, its recall is increased by 0.8% and 1.2% while ${F}_1$ -parameter is improved by 0.6% and 0.8%, respectively, in the two test sites.
21 June 2023 in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
With the advancement of remote sensing technology, various new satellite sensors have emerged as the times require. Remote sensing images acquired by different sensors exhibit different characteristics due to their distinct imaging mechanisms. The fusion of Synthetic Aperture Radar (SAR) and optical remote sensing images is valuable for specific remote sensing image applications, as it enables the extraction of texture features from SAR images while preserving the spectral information of optical images. Several existing fusion approaches have been proposed in recent years, including the Nonsubsampled Shearlet Transform Pulse Coupled Neural Network (NSST-PCNN), which is a typical and effective fusion method. However, it suffers from the inconsistency in regional edge information due to the lack of target fusion rules. To address this issue, we propose a new method called MS-NSST-PCNN for multi-model fusion of SAR and optical remote sensing images. This method incorporates the multiScale morphological gradient (MSMG) into NSST-PCNN, which is a technique used to detect edges and enhance the utilization of edge characteristics. The fusion results of two polarization modes, VV and VH are evaluated in combination with existing methods, using image fusion accuracy and visual interpretation criteria. The results demonstrate that for Sentinel 1 and Landsat 8 OLI image fusion the proposed MS-NSST-PCNN method achieves higher correlation coefficients and lower spectral distortion with VV polarization compared to traditional methods in two study areas. Moreover, the proposed method also exhibits better performance for higher spatial resolution GF3 and GF2 images. In subsequent applications of land feature classification, the fusion results of the proposed method achieve higher accuracy than those of other fusion methods or source images applied directly. In the urban and rural application scenarios, the overall classification accuracy of the fusion results can reach 0.87 and 0.88, respectively, which has increased by 8.75% and 23.94% compared with that of using the Landsat8 OLI source images directly.