ABSTRACT This study aimed to evaluate the dynamics driving seasonal and interannual surface chlorophyll-a (chl-a) variability in the Karimata Strait (KS). The analysis shows that high chl-a concentrations were observed in the KS during December – February (northwest monsoon season) extending northwestward along the eastern coast of Peninsular Malaysia and northeastward along the northern coast of Borneo Island. This high chl-a concentration contrasts with the low chl-a concentration observed along the southern coast of Sumatra and Java during the same season. A substantial upwelling signal was identified in the central-eastern section of the South China Sea and along the northern coast of Java and the eastern coast of Sumatra from December to February. The positive wind stress curl forced downwelling in the southern section of the KS and the northern half of the Java Sea off the coast of Kalimantan. The elevated chl-a concentration in the KS during the northwest monsoon season could not be attributed to wind dynamics alone. The increase in surface chl-a concentration during the northwest monsoon season was associated with an increase in allochthonous nutrients from river discharges caused by increased precipitation over land. On an interannual timescale, a high chl-a concentration was observed during the southeast monsoon season when the La Niña event occurred in the tropical Pacific. The warm SST, associated with favourable downwelling winds, was observed over the entire the KS. It is suggested that anomalous chl-a that bloomed during the La Niña event was associated with anomalously high precipitation over land that transported nutrients to KS through river discharge.

ABSTRACT Target segmentation of remote sensing images has always been a hotspot in image processing. This paper proposes a new semantic segmentation technology for remote sensing images, which uses Unet as the backbone and combines attention mechanism and feature enhancement module. The feature enhancement module can enlarge the information of the region of interest (ROI) to improve the contrast of the image; the attention mechanism includes spatial and channel attention modules, which can obtain more detailed information of the desired target while suppressing other useless information. This paper improves the loss function of the traditional Unet. On the basis of the sparse categorical cross-entropy function, the mean squared logarithmic error function is added, which can effectively improve the accuracy of semantic segmentation. The experimental results show that the algorithm has higher computational accuracy than Unet, DeepLabV3, SegNet, PSPNet, CBAM and DAnet while having the computational speed of FCN and Unet in model testing and validation.

ABSTRACT The Daning River Basin is a typical representative of the ‘mountain forest’ in the Three Gorges Reservoir (TGR) area of the Yangtze River. In recent years, with the completion of the Three Gorges Project, the local vegetation has degraded, soil erosion has become severe, and there is an urgent need to assess the environmental quality. Data were fused using the MODIS Normalized Difference Vegetation Index (NDVI) (250 m) and Landsat NDVI (30 m) through an Enhanced Spatial and Temporal Adaptive Reflectance Fusion Model (ESTARFM) to obtain ESTARFM NDVI (30 m). This data was then entered into the Carnegie Ames Stanford Approach (CASA) model, along with meteorological and land use data, to calculate vegetation net primary productivity (NPP) and trends in the Daning River Basin. The detection of drivers with a high impact on vegetation NPP was done using a Geodetector. The results show that: (1) In terms of spatial and temporal changes, the annual average NPP of the watershed during the 13 years from 2008 to 2020 generally showed an upward trend, with the average yearly vegetation NPP being 512.33gC·m−2·a−1, exhibiting a low southwest to surrounding increasing trend along the river. (2) The spatial and temporal variation of vegetation NPP is influenced by several factors synergistically, with elevation, temperature, and distance from settlements being the dominant factors. The interaction of these two factors can enhance the explanatory power of vegetation NPP. Through the estimation of vegetation NPP and the analysis of influencing factors in the Daning River Basin, this study can provide a reference for the ecological restoration and management of vegetation NPP in small watersheds under similar environments.

ABSTRACT Changes in Vegetation Indices (VIs) over time can describe vegetation phenology; however, it is not known which phenological features contribute most to land cover classification. Feature selection could potentially solve this problem. In this study, phenological feature importance and selection were evaluated by using two-year Sentinel-2 (S-2) data and single-date PlanetScope (PS) to classify a 50 km2 native podocarp forest in New Zealand. The study area was classified into nine classes. Single-date PS and S-2 data were fused to a base image with the same spatial resolution as PS and 8 bands containing spectral data from S-2; this image was used to produce 30 Vegetation Indices (VIs). Phenological features – amplitude (AMP) and phase (PH) were extracted from these VIs using time-series S-2 only, and harmonic analysis in Google Earth Engine. For accurately classifying forests and identifying the most important features, three classification scenarios (fused bands & VIs, fused bands & phenological features, fused bands & VIs & phenological features) were developed using a Neural Network. Variable Selection Using Random Forest (VSURF) was applied on these scenarios to evaluate the impact of feature selection. Results indicate that VSURF could reduce the time needed for the classification while maintaining a comparable level of accuracy. Phenological features improved accuracy from 90% to 94%, driven mostly by Red-Edge Triangulated Vegetation Index-AMP&PH, Normalised Near-Infrared-PH, Greenness Index-PH, Water Body Index-PH, Normalised Difference Vegetation Index-PH, Normalised Green-PH, Red-Edge Normalised Difference Vegetation Index-PH, Leaf Chlorophyll Content-AMP, and Simple Near-Infrared and Blue Ratio-PH. These features reflect changes in the structure, biochemical, and physiological characteristics of the canopy. A lack of ground-based measurements precluded an evaluation of the accuracy of these phenological aspects and an explanation of their distinctive contribution to the model. Overall, the findings show that specific phenological features can improve the classification of New Zealand’s indigenous podocarp forests.

ABSTRACT Land Use/Land Cover (LULC) classification has become increasingly important in various fields, including ecological and environmental protection, urban planning, and geological disaster monitoring. With the development of high-resolution remote sensing satellite technology, there is a growing focus on achieving precise LULC classification. However, the accuracy of fine-grained LULC classification is challenged by the high intra-class diversity and low inter-class separability inherent in high-resolution remote sensing images. To address this challenge, this paper proposes a novel multi-path feature fusion semantic segmentation model, called MPFFNet, which combines the segmentation results of convolutional neural networks with traditional filtering processes to achieve finer LULC classification. MPFFNet consists of three modules: the Improved Encoder Module (IEM) extracts contextual and spatial detail information through the backbone network, DASPP, and MFEAM; the Improved Decoder Module (IDM) utilizes the Cascade Feature Fusion (CFF) module to effectively merge shallow and deep information; and the Feature Fusion Module (FAM) enables dual-path feature fusion using a convolutional neural network and Gabor Filter. Experimental results on the large-scale classification set and the fine land-cover classification set of the Gaofen Image Dataset (GID) demonstrate the effectiveness of the proposed method, achieving mIoU scores of 81.02% and 77.83%, respectively. These scores outperform U-Net by 7.95% and 3.28%, respectively. Therefore, we believe that our model can deliver superior results in the task of LULC classification.

ABSTRACT Illicit maritime activity, such as piracy and smuggling, is a global issue that often occurs in areas where interdiction authorities are sparse, ground-based monitoring technologies like radar lack range and coverage, and dark vessels abound. Our objective was to assess vessel congregation patterns, identify likely beaching areas based on terrain maps, and use them to narrow the search for potential smuggling transfer and overland routes in a specific region along the Puntland coast of Somalia. To accomplish this goal, we developed automated protocols applied to WorldView satellite imagery for (1) vessel detection and size classification, (2) shallow-water bathymetric characterization, and (3) coastal topography mapping. Utilizing a single sensor for vessel detection and topographic and bathymetric extraction at high spatial resolution and at a high re-visit rate provides a simplification to near-shore characterization for monitoring purposes. The extracted topography and bathymetry are then presented as a single, comprehensive perspective of a coastal region. The vessel-detection algorithm identified all vessels larger than approximately 15 metres in length (35 of 35), but misidentified six artefacts (i.e. false positives), resulting in an overall accuracy of 85%. The combined vessel and terrain maps facilitated the identification of a potential beaching and overland transportation route location.

ABSTRACT LiDAR and camera are pivotal sensors of 3D (three-dimensional) object detection. As a result of their different characteristics, increasingly multimodal-based object detection methods have been proposed. Now, popular methods are to hardly associate camera features with LiDAR features, but the features are frequently enhanced and aggregated, so there is a major challenge in how to align two features effectively. Therefore, we propose CAF-RCNN. On the basis of PointRCNN, using Feature Pyramid Network (FPN) to extract advanced semantic features at different scales, then fusing these features with the LiDAR features of the Set ion (SA) module output in PointRCNN and subsequent steps. Regarding the features fusion module, we design a module based on the cross-attention mechanism, CAFM (Cross-Attention Fusion Module). It combines two channel attention streams in a cross-over fashion to utilize rich details about significant objects in the Image Stream and Geometric Stream. We did a lot of experiments on the KITTI dataset, and the result shows that our method is 6.43% higher than PointRCNN in 3D accuracy.

ABSTRACT The fine 3D model is the essential spatial information for the construction of a smart city. UAV aerial images with large-scale scene perception ability are common data sources for 3D modelling of cities at present. However, in some complex urban areas, a single aerial image is difficult to capture the 3D scene information because of the existence of some problems such as inaccurate edges, holes, and blurred building facade textures due to changes in perspective and area occlusion. Therefore, how to solve perspective changes and area occlusion of the aerial image quickly and efficiently has become an important problem. The ground image can be used as an important supplement to solve the problem of missing bottom and area occlusion in oblique photography modelling. Thus, this article proposes a progressive matching method via multi-scale context feature coding network to achieve robust matching of aerial-ground remote sensing images, which provides better technical support for urban modelling. The main idea consists of three parts: (1) a multi-scale context feature coding network is designed to extract feature on aerial-ground images efficiently; (2) a block-based matching strategy is proposed to pay more attention to local features of the aerial-ground images; (3) a progressive matching method is applied in block matching stage to obtain more accurate features. We used eight sets of typical data, such as aerial images captured by the drone DJI-MAVIC2 and ground images captured by handheld devices as experimental objects, and compared them with algorithms such as SIFT, D2-net, DFM and SuperGlue. Experimental results show that our proposed aerial-ground image matching method has a good performance that the average NCM has improved 2.1–8.2 times, and the average rate of correct matching has an average increase of 26% points with the average root of mean square error is only 1.48 pixels.

ABSTRACT According to the World Bank data catalogue, records of reservoirs and their associated dams are present summing up a capacity of km3 of water. They play a crucial role in providing potable and irrigation water and, therefore, it is of paramount interest to effectively monitor such critical infrastructures. An effective approach is based on satellite remote sensing and, in particular, on the Synthetic Aperture Radar (SAR). In this paper, we critically investigate the use of polarimetric SAR measurements for reservoirs’ waterline estimation. Measurements of the novel COSMO-SkyMed Second Generation (CSG) X-band quad-polarimetric SAR related to the San Giuliano reservoir, in the South of Italy, are used to carry out an electromagnetic analysis of the different polarimetric scattering returns. Experimental results show that the cross-polarized channel, as well as the inter-channel phase, are noisy and, therefore, uninformative when used to design coherent polarimetric waterline extraction methods. From an electromagnetic viewpoint, this is due to the peculiarities of the reservoirs that call for low surface roughness and negligible wave pattern that, at once, result in a joint combination of un-tilted Bragg scattering and specular reflection. This implies that a low co-polarized backscatter and a cross-polarized signal largely below the system noise floor are to be expected. As a consequence, waterline extraction approaches that do not exploit the inter-channel phase, the so-called incoherent approaches, are shown to outperform the coherent ones.

ABSTRACT How to remove various kinds of noises of hyperspectral images (HSIs) simultaneously is an important issue that we face, which will directly affect the subsequent application of HSIs. Many HSI denoising techniques have been proposed for this purpose and deep convolutional neural network (CNN) is the most effective approach in recent years. However, existing HSI denoising methods are usually unsatisfactory in removing mixed noises and preserving more details. In this paper, we propose a novel noise intensity estimation guided progressive network (NIEG-PNet) to address these problems. To be specific, we design three core modules to improve the HSI denoising accuracy. First, we propose a noise intensity estimation module (NIEM) to capture multiscale noises and structural characteristics of HSI, which can also be a noise prior to guide the network learning. Second, we propose a recurrent feedback grouped denoising module (RFGDM) to fully capture the spectral correlation. Moreover, we design a dual self-attention module including a position self-attention module and a channel self-attention module to exploit the local and global spatial and spectral correlations and alleviate the problem of small available training HSIs. Last, we propose a global spatial-spectral consistency module (GSSCM), which designs a new parallel structure to combine the two-dimensional and the three-dimensional convolutions more effectively. Moreover, it can explore the relationship between the spectral and the horizontal or vertical direction of the spaces. The experimental results on both synthetic and real-data experiments show the superiority of the proposed NIEG-PNet compared to other traditional and advanced HSI denoising methods.