Search

Salient Object Detection (SOD) is gradually applied in natural scene images. However, due to the apparent differences between optical remote sensing images and natural scene images, directly applying the SOD of natural scene images to optical remote sensing images has limited performance in global context information. Therefore, salient object detection in optical remote sensing images (ORSI-SOD) is challenging. Optical remote sensing images usually have large-scale variations. However, the vast majority of networks are based on Convolutional Neural Network (CNN) backbone networks such as VGG and ResNet, which can only extract local features. To address this problem, we designed a new model that employs a transformer-based backbone network capable of extracting global information and remote dependencies. A new framework is proposed for this question, named Global and Multiscale Aggregate Network for Saliency Object Detection in Optical Remote Sensing Images (GMANet). In this framework, the Pyramid Vision Transformer (PVT) is an encoder to catch remote dependencies. A Multiscale Attention Module (MAM) is introduced for extracting multiscale information. Meanwhile, a Global Guiled Brach (GGB) is used to learn the global context information and obtain the complete structure. Four MAMs are densely connected to this GGB. The Aggregate Refinement Module (ARM) is used to enrich the details of edge and low-level features. The ARM fuses global context information and encoder multilevel features to complement the details while the structure is complete. Extensive experiments on two public datasets show that our proposed framework GMANet outperforms 28 state-of-the-art methods on six evaluation metrics, especially E-measure and F-measure. It is because we apply a coarse-to-fine strategy to merge global context information and multiscale information. [ABSTRACT FROM AUTHOR]

Salient Object Detection (SOD) is gradually applied in natural scene images. However, due to the apparent differences between optical remote sensing images and natural scene images, directly applying the SOD of natural scene images to optical remote sensing images has limited performance in global context information. Therefore, salient object detection in optical remote sensing images (ORSI-SOD) is challenging. Optical remote sensing images usually have large-scale variations. However, the vast majority of networks are based on Convolutional Neural Network (CNN) backbone networks such as VGG and ResNet, which can only extract local features. To address this problem, we designed a new model that employs a transformer-based backbone network capable of extracting global information and remote dependencies. A new framework is proposed for this question, named Global and Multiscale Aggregate Network for Saliency Object Detection in Optical Remote Sensing Images (GMANet). In this framework, the Pyramid Vision Transformer (PVT) is an encoder to catch remote dependencies. A Multiscale Attention Module (MAM) is introduced for extracting multiscale information. Meanwhile, a Global Guiled Brach (GGB) is used to learn the global context information and obtain the complete structure. Four MAMs are densely connected to this GGB. The Aggregate Refinement Module (ARM) is used to enrich the details of edge and low-level features. The ARM fuses global context information and encoder multilevel features to complement the details while the structure is complete. Extensive experiments on two public datasets show that our proposed framework GMANet outperforms 28 state-of-the-art methods on six evaluation metrics, especially E-measure and F-measure. It is because we apply a coarse-to-fine strategy to merge global context information and multiscale information.

Time series remote sensing products with both fine spatial and dense temporal resolutions are urgently needed for many earth system studies. The development of small satellite constellations with identical sensors affords novel opportunities to provide such kind of earth observations. In this paper, a new dense time series 30-m image product was proposed respectively based on an 8-day, 16-day and monthly composition. The products were composited by the Charge Coupled Device (CCD) images from the 2-day revisit small satellite constellation for environmental monitoring and disaster mitigation of China (HJ-1A/B). Taking the Zoige plateau in China as a case area where it is covered by highly heterogeneous vegetation landscapes, a detailed methodology was introduced on how to use 183 scenes of CCD images in 2010 to create composite products. The quality of the HJ CCD composites was evaluated by inter-comparison with the monthly 30-m global Web-Enabled Landsat Data (WELD), 16-day 500-m MODIS NDVI, and 8-day 500-m MODIS surface reflectance products. Results showed that the radiometric consistency between HJ and WELD composited Top Of Atmosphere (TOA) reflectance was in good agreement except for May, June, July and August when more clouds and invalid data gaps appeared in WELD. Visual assessment and temporal profile analysis also revealed that HJ possessed better visual effects and temporal coherence than that of WELD. The comparison between HJ and MODIS products indicated that HJ composites were radiometrically consistent with MODIS products over areas consisting of large patches of homogeneous surface types, but can better reflect the detailed spatial differences in regions with heterogeneous landscapes. This paper highlights the potential of compositing HJ-1A/B CCD images, allowing for providing a cloud free, time-space consistent, 30-m spatial resolution, and dense in time series image product. Meanwhile, the proposed products could also be treated as a reference to generate regional or even global composited products for the on-going satellite constellations and even for the forthcoming satellite missions such as Sentinel-2A/B.

Time series remote sensing products with both fine spatial and dense temporal resolutions are urgently needed for many earth system studies. The development of small satellite constellations with identical sensors affords novel opportunities to provide such kind of earth observations. In this paper, a new dense time series 30-m image product was proposed respectively based on an 8-day, 16-day and monthly composition. The products were composited by the Charge Coupled Device (CCD) images from the 2-day revisit small satellite constellation for environmental monitoring and disaster mitigation of China (HJ-1A/B). Taking the Zoige plateau in China as a case area where it is covered by highly heterogeneous vegetation landscapes, a detailed methodology was introduced on how to use 183 scenes of CCD images in 2010 to create composite products. The quality of the HJ CCD composites was evaluated by inter-comparison with the monthly 30-m global Web-Enabled Landsat Data (WELD), 16-day 500-m MODIS NDVI, and 8-day 500-m MODIS surface reflectance products. Results showed that the radiometric consistency between HJ and WELD composited Top Of Atmosphere (TOA) reflectance was in good agreement except for May, June, July and August when more clouds and invalid data gaps appeared in WELD. Visual assessment and temporal profile analysis also revealed that HJ possessed better visual effects and temporal coherence than that of WELD. The comparison between HJ and MODIS products indicated that HJ composites were radiometrically consistent with MODIS products over areas consisting of large patches of homogeneous surface types, but can better reflect the detailed spatial differences in regions with heterogeneous landscapes. This paper highlights the potential of compositing HJ-1A/B CCD images, allowing for providing a cloud free, time-space consistent, 30-m spatial resolution, and dense in time series image product. Meanwhile, the proposed products could also be treated as a reference to generate regional or even global composited products for the on-going satellite constellations and even for the forthcoming satellite missions such as Sentinel-2A/B. [ABSTRACT FROM AUTHOR]