Search Results (379)

Existing convolution neural network (CNN)-based video synthetic aperture radar (SAR) moving target shadow detectors are difficult to model long-range dependencies, while transformer-based ones often suffer from greater complexity. To handle these issues, this paper proposes MambaShadowDet, a novel lightweight deep learning (DL) detector based on a state space model (SSM), dedicated to high-speed and high-accuracy moving target shadow detection in video SAR images. By introducing SSM with the linear complexity into YOLOv8, MambaShadowDet effectively captures the global feature dependencies while relieving computational load. Specifically, it designs Mamba-Backbone, combining SSM and CNN to effectively extract both global contextual and local spatial information, as well as a slim path aggregation feature pyramid network (Slim-PAFPN) to enhance multi-level feature extraction and further reduce complexity. Abundant experiments on the Sandia National Laboratories (SNL) video SAR data show that MambaShadowDet achieves superior moving target shadow detection performance with a detection accuracy of 80.32% F1 score and an inference speed of 44.44 frames per second (FPS), outperforming existing models in both accuracy and speed. Full article

Lane detection is a pivotal technology of the intelligent driving system. By identifying the position and shape of the lane, the vehicle can stay in the correct lane and avoid accidents. Image-based deep learning is currently the most advanced method for lane detection. Models using this method already have a very good recognition ability in general daytime scenes, and can almost achieve real-time detection. However, these models often fail to accurately identify lanes in challenging scenarios such as night, dazzle, or shadows. Furthermore, the lack of diversity in the training data restricts the capacity of the models to handle different environments. This paper proposes a novel method to train CycleGAN with existing daytime and nighttime datasets. This method can extract features of different styles and multi-scales, thereby increasing the richness of model input. We use CycleGAN as a domain adaptation model combined with an image segmentation model to boost the model’s performance in different styles of scenes. The proposed consistent loss function is employed to mitigate performance disparities of the model in different scenarios. Experimental results indicate that our method enhances the detection performance of original lane detection models in challenging scenarios. This research helps improve the dependability and robustness of intelligent driving systems, ultimately making roads safer and enhancing the driving experience. Full article

Forecasting stock markets is challenging due to the influence of various internal and external factors compounded by the effects of globalization. This study introduces a data-driven approach to forecast S&P 500 returns by incorporating macroeconomic indicators including gold and oil prices, the volatility index, economic policy uncertainty, the financial stress index, geopolitical risk, and shadow short rate, with ten technical indicators. We propose three hybrid deep learning models that sequentially combine convolutional and recurrent neural networks for improved feature extraction and predictive accuracy. These models include the deep belief network with gated recurrent units, the LeNet architecture with gated recurrent units, and the LeNet architecture combined with highway networks. The results demonstrate that the proposed hybrid models achieve higher forecasting accuracy than the single deep learning models. This outcome is attributed to the complementary strengths of convolutional networks in feature extraction and recurrent networks in pattern recognition. Additionally, an analysis using the Shapley method identifies the volatility index, financial stress index, and economic policy uncertainty as the most significant predictors, underscoring the effectiveness of our data-driven approach. These findings highlight the substantial impact of contemporary uncertainty factors on stock markets, emphasizing their importance in studies analyzing market behaviour. Full article

One of the most important nutrients needed for fruit tree growth is nitrogen. For orchards to get targeted, well-informed nitrogen fertilizer, accurate, large-scale, real-time monitoring, and assessment of nitrogen nutrition is essential. This study examines the Leaf Nitrogen Content (LNC) of the custard apple tree, a noteworthy fruit tree that is extensively grown in China’s Yunnan Province. This study uses an ensemble learning technique based on multiple machine learning algorithms to effectively and precisely monitor the leaf nitrogen content in the tree canopy using multispectral canopy footage of custard apple trees taken via Unmanned Aerial Vehicle (UAV) across different growth phases. First, canopy shadows and background noise from the soil are removed from the UAV imagery by using spectral shadow indices across growth phases. The noise-filtered imagery is then used to extract a number of vegetation indices (VIs) and textural features (TFs). Correlation analysis is then used to determine which features are most pertinent for LNC estimation. A two-layer ensemble model is built to quantitatively estimate leaf nitrogen using the stacking ensemble learning (Stacking) principles. Random Forest (RF), Adaptive Boosting (ADA), Gradient Boosting Decision Trees (GBDT), Linear Regression (LR), and Extremely Randomized Trees (ERT) are among the basis estimators that are integrated in the first layer. By detecting and eliminating redundancy among base estimators, the Least Absolute Shrinkage and Selection Operator regression (Lasso)model used in the second layer improves nitrogen estimation. According to the analysis results, Lasso successfully finds redundant base estimators in the suggested ensemble learning approach, which yields the maximum estimation accuracy for the nitrogen content of custard apple trees’ leaves. With a root mean square error (RMSE) of 0.059 and a mean absolute error (MAE) of 0.193, the coefficient of determination (R²) came to 0. 661. The significant potential of UAV-based ensemble learning techniques for tracking nitrogen nutrition in custard apple leaves is highlighted by this work. Additionally, the approaches investigated might offer insightful information and a point of reference for UAV remote sensing applications in nitrogen nutrition monitoring for other crops. Full article

Ethnic minority architecture is a vital carrier of the cultural heritage of ethnic minorities in China, and its quick and accurate extraction from remote sensing images is highly important for promoting the application of remote sensing information in urban management and architectural heritage protection. Taking Buyi architecture in China as an example, this paper proposes a minority architectural heritage identification method that combines low-altitude unmanned aerial vehicle (UAV) remote sensing technology and an improved deep learning algorithm. First, UAV images are used as the data source to provide high-resolution images for research on ethnic architecture recognition and to solve the problems associated with the high costs, time consumption, and destructiveness of traditional methods for ethnic architecture recognition. Second, to address the lack of edge pixel features in the sample images and reduce repeated labeling of the same sample, the ethnic architecture in entire remote sensing images is labeled on the Arcgis platform, and the sliding window method is used to cut the image data and the corresponding label file with a 10% overlap rate. Finally, an attention mechanism SE module is introduced to improve the DeepLabV3+ network model structure and achieve superior ethnic building recognition results. The experimental data fully show that the model’s accuracy reaches as high as 0.9831, with an excellent recall rate of 0.9743. Moreover, the F1 score is stable at a high level of 0.9787, which highlights the excellent performance of the model in terms of comprehensive evaluation indicators. Additionally, the intersection/union ratio (IoU) of the model is 0.9582, which further verifies its high precision in pixel-level recognition tasks. According to an in-depth comparative analysis, the innovative method proposed in this paper solves the problem of insufficient feature extraction of sample edge pixels and substantially reduces interference from complex environmental factors such as roads, building shadows, and vegetation with the recognition results for ethnic architecture. This breakthrough greatly improves the accuracy and robustness of the identification of architecture in low-altitude remote sensing images and provides strong technical support for the protection and intelligent analysis of architectural heritage. Full article

Shadows degrade image quality and complicate interpretation, underscoring the importance of accurate shadow detection for many image analysis tasks. However, due to the complex backgrounds and variable shadow characteristics of remote sensing images (RSIs), existing methods often struggle with accurately detecting shadows of various scales and misclassifying dark, non-shaded areas as shadows. To address these issues, we proposed a comprehensive shadow detection network called MAMNet. Firstly, we proposed a multi-scale spatial channel attention fusion module, which extracted multi-scale features incorporating both spatial and channel information, allowing the model to flexibly adapt to shadows of different scales. Secondly, to address the issue of false detection in non-shadow areas, we introduced a criss-cross attention module, enabling non-shadow pixels to be compared with other shadow and non-shadow pixels in the same row and column, learning similar features of pixels in the same category, which improved the classification accuracy of non-shadow pixels. Finally, to address the issue of important information from the other two modules being lost due to continuous upsampling during the decoding phase, we proposed an auxiliary branch module to assist the main branch in decision-making, ensuring that the final output retained the key information from all stages. The experimental results demonstrated that the model outperformed the current state-of-the-art RSI shadow detection method on the aerial imagery dataset for shadow detection (AISD). The model achieved an overall accuracy (OA) of 97.50%, an $F_1$ score of 94.07%, an intersection over union (IOU) of 88.87%, a precision of 95.06%, and a BER of 4.05%, respectively. Additionally, visualization results indicated that our model could effectively detect shadows of various scales while avoiding false detection in non-shadow areas. Therefore, this model offers an efficient solution for shadow detection in aerial imagery. Full article

Evaluating satellite water extraction indices, particularly for coastal environments, guarantees that satellite-derived water maps are as accurate and functional as possible, notwithstanding the unique complexities these areas present. Variability in salinity levels, intricate land-water boundaries, dynamic sediment loads, and tidal fluctuations often complicate coastal water mapping. Sandwip Island in Bangladesh is one of the most complex and dynamic coastal environments in the world and is our area of focus. Six water information extraction indices were evaluated: normalized-difference vegetation index (NDVI), modified normalized difference water index (MNDWI), automated water extraction index for built-up areas (AWEI_nsh) and shadows (AWEI_sh), multi-band water index (MBWI), and normalized difference water index (NDWI), using Sandwip Island’s satellite Landsat imagery acquired in February 1990, 2000, 2010, and 2020. The results showed that NDWI performed the best based on the total area obtained and classification accuracy. NDWI was then used to assess the erosion and accretion dynamics of the island for the study period (1990–2020). In the period 1990–2000, the island saw significant erosion and accretion along its coastlines in all parts, while the 2000–2010 period indicated that the island eroded on all sides. However, the situation was totally opposite during 2010–2020. The results illustrated the best performance of the NDWI algorithm in mapping surface water in the complex and dynamic Sandwip coastal environment. Also, erosion and accretion change temporally and spatially on the island. While this study is confined to Sandwip Island in Bangladesh, the findings hold the potential for broader applicability in regions with comparable characteristics. Full article

The combustion efficiency of wood pellets is partly affected by their average length. The ISO 17829 standard defines the methodology for assessing the average length of sample pellets, but the method does not always lead to representative data. Furthermore, a standard analysis is time-consuming as it requires manual measurement of the pellets using a caliper. This paper, whilst evaluating the effect of pellet length on combustion efficiency, proposes a pending-patented dimensional image processing method (DIP) for assessing pellet length. DIP allows the dimensional data of grouped and stacked pellets to be obtained by exploiting the shadows produced by pellets when exposed to a light source, assuming that different-sized pellets produce different shadows. Thus, the proposed method allows for the extraction of dimensional information from non-distinct objects, overcoming the reliance of classical image processing methods on object distance for effective segmentation. Combustion tests, carried out using pellets varying only in length, confirmed the influence of length on combustion efficiency. Shorter pellets, compared to longer ones, significantly reduced CO emissions by up to 94% (mg/MJ). However, they exhibited a higher fuel mass consumption rate (kg/h), with an increase of up to 22.8% compared to the longest sample. In addition, longer pellets produced fewer but larger shadows than shorter ones. Further studies are needed to correlate the number and size of shadows with samples’ average length so that DIP could be implemented in stoves and programmed to communicate with the control unit and automatically optimize the setting in order to improve combustion efficiency. Full article

Effective detection of the contours of cloud masks and estimation of their distribution can be of practical help in studying weather changes and natural disasters. Existing deep learning methods are unable to extract the edges of clouds and backgrounds in a refined manner when detecting cloud masks (shadows) due to their unpredictable patterns, and they are also unable to accurately identify small targets such as thin and broken clouds. For these problems, we propose MDU-Net, a multiscale dual up-sampling segmentation network based on an encoder–decoder–decoder. The model uses an improved residual module to capture the multi-scale features of clouds more effectively. MDU-Net first extracts the feature maps using four residual modules at different scales, and then sends them to the context information full flow module for the first up-sampling. This operation refines the edges of clouds and shadows, enhancing the detection performance. Subsequently, the second up-sampling module concatenates feature map channels to fuse contextual spatial information, which effectively reduces the false detection rate of unpredictable targets hidden in cloud shadows. On a self-made cloud and cloud shadow dataset based on the Landsat8 satellite, MDU-Net achieves scores of 95.61% in PA and 84.97% in MIOU, outperforming other models in both metrics and result images. Additionally, we conduct experiments to test the model’s generalization capability on the landcover.ai dataset to show that it also achieves excellent performance in the visualization results. Full article

Automatic large-scale building extraction from the LiDAR point clouds and remote sensing images is a growing focus in the fields of the sensor applications and remote sensing. However, this building extraction task remains highly challenging due to the complexity of building sizes, shapes, and surrounding environments. In addition, the discreteness, sparsity, and irregular distribution of point clouds, lighting, and shadows, as well as occlusions of the images, also seriously affect the accuracy of building extraction. To address the above issues, we propose a new unsupervised building extraction algorithm PBEA (Point and Pixel Building Extraction Algorithm) based on a new dual P-snake model (Dual Point and Pixel Snake Model). The proposed dual P-snake model is an enhanced active boundary model, which uses both point clouds and images simultaneously to obtain the inner and outer boundaries. The proposed dual P-snake model enables interaction and convergence between the inner and outer boundaries to improve the performance of building boundary detection, especially in complex scenes. Using the dual P-snake model and polygonization, this proposed PBEA can accurately extract large-scale buildings. We evaluated our PBEA and dual P-snake model on the ISPRS Vaihingen dataset and the Toronto dataset. The experimental results show that our PBEA achieves an area-based quality evaluation metric of 90.0% on the Vaihingen dataset and achieves the area-based quality evaluation metric of 92.4% on the Toronto dataset. Compared with other methods, our method demonstrates satisfactory performance. Full article

The semantic segmentation of high-resolution remote sensing images (HRRSIs) faces persistent challenges in handling complex architectural structures and shadow occlusions, limiting the effectiveness of existing deep learning approaches. To address these limitations, we propose an attention-focused feature enhancement network (AFENet) with a novel encoder–decoder architecture. The encoder architecture combines ResNet50 with a parallel multistage feature enhancement group (PMFEG), enabling robust feature extraction through optimized channel reduction, scale expansion, and channel reassignment operations. Building upon this foundation, we develop a global multi-scale attention mechanism (GMAM) in the decoder that effectively synthesizes spatial information across multiple scales by learning comprehensive global–local relationships. The architecture is further enhanced by an efficient feature-weighted fusion module (FWFM) that systematically integrates remote spatial features with local semantic information to improve segmentation accuracy. Experimental results across diverse scenarios demonstrate that AFENet achieves superior performance in building structure detection, exhibiting enhanced segmentation connectivity and completeness compared to state-of-the-art methods. Full article

The aim of this article is to analyse the global environmental impact of wind farms, i.e., the effects on human health and the local ecosystem. Compared to conventional energy sources, wind turbines emit significantly fewer greenhouse gases, which helps to mitigate global warming. During the life cycle of a wind farm, 86% of CO₂ emissions are generated by the extraction of raw materials and the manufacture of wind turbine components. The water consumption of wind farms is extremely low. In the operational phase, it is 4 L/MWh, and in the life cycle, one water footprint is only 670 L/MWh. However, wind farms occupy a relatively large total area of 0.345 ± 0.224 km²/MW of installed capacity on average. For this reason, wind farms will occupy more than 10% of the land area in some EU countries by 2030. The impact of wind farms on human health is mainly reflected in noise and shadow flicker, which can cause insomnia, headaches and various other problems. Ice flying off the rotor blades is not mentioned as a problem. On a positive note, the use of wind turbines instead of conventionally operated power plants helps to reduce the emission of particulate matter 2.5 microns or less in diameter (PM 2.5), which are a major problem for human health. In addition, the non-carcinogenic toxicity potential of wind turbines for humans over the entire life cycle is one of the lowest for energy plants. Wind farms can have a relatively large impact on the ecological system and biodiversity. The destruction of animal migration routes and habitats, the death of birds and bats in collisions with wind farms and the negative effects of wind farm noise on wildlife are examples of these impacts. The installation of a wind turbine at sea generates a lot of noise, which can have a significant impact on some marine animals. For this reason, planners should include noise mitigation measures when selecting the site for the future wind farm. The end of a wind turbine’s service life is not a major environmental issue. Most components of a wind turbine can be easily recycled and the biggest challenge is the rotor blades due to the composite materials used. Full article

Alpine lakes located in high-altitude mountainous regions act as vital sentinels of environmental change. Remote-sensing-based identification of these lakes is crucial for understanding their response to climate variations and for assessing associated disaster risks. However, the complex terrain and weather conditions in these areas pose significant challenges to accurate detection. This paper proposes a method that leverages the high precision of deep learning for small lake and lake boundary extraction combined with deep learning to eliminate noise and errors in the identification results. Using Sentinel-2 data, we accurately identified and delineated alpine lakes in the eastern Himalayas. A total of 2123 lakes were detected, with an average lake area of 0.035 km². Notably, 76% of these lakes had areas smaller than 0.01 km². The slope data is crucial for the lake classification model in eliminating shadow noise. The accuracy of the proposed lake classification model reached 97.7%. In the identification of small alpine lakes, the recognition rate of this method was 96.4%, significantly surpassing that of traditional deep learning approaches. Additionally, this method effectively eliminated most shadow noise present in water body detection results obtained through machine learning techniques. Full article

Vegetation segmentation plays a crucial role in accurately monitoring and analyzing vegetation cover, growth patterns, and changes over time, which in turn contributes to environmental studies, land management, and assessing the impact of climate change. This study explores the potential of a multi-scale convolutional neural network (MSCNN) design for object classification, specifically focusing on vegetation detection. The MSCNN is designed to integrate multi-scale feature extraction and attention mechanisms, enabling the model to capture both fine and coarse vegetation patterns effectively. Moreover, the MSCNN architecture integrates multiple convolutional layers with varying kernel sizes (3 × 3, 5 × 5, and 7 × 7), enabling the model to extract features at different scales, which is vital for identifying diverse vegetation patterns across various landscapes. Vegetation detection is demonstrated using three diverse datasets: the CamVid dataset, the FloodNet dataset, and the multispectral RIT-18 dataset. These datasets present a range of challenges, including variations in illumination, the presence of shadows, occlusion, scale differences, and cluttered backgrounds, which are common in real-world scenarios. The MSCNN architecture allows for the integration of information from multiple scales, facilitating the detection of diverse vegetation types under varying conditions. The performance of the proposed MSCNN method is rigorously evaluated and compared against state-of-the-art techniques in the field. Comprehensive experiments showcase the effectiveness of the approach, highlighting its robustness in accurately segmenting and classifying vegetation even in complex environments. The results indicate that the MSCNN design significantly outperforms traditional methods, achieving a remarkable global accuracy and boundary F1 score (BF score) of up to 98%. This superior performance underscores the MSCNN’s capability to enhance vegetation detection in imagery, making it a promising tool for applications in environmental monitoring and land use management. Full article

The Lunarminer framework explores the use of biomimetic swarm robotics, inspired by the division of labor in leafcutter ants and the synchronized flashing of fireflies, to enhance lunar water ice extraction. Simulations of water ice extraction within Shackleton Crater showed that the framework may improve task allocation, by reducing the extraction time by up to 40% and energy consumption by 31% in scenarios with high ore block quantities. This system, capable of producing up to 181 L of water per day from excavated regolith with a conversion efficiency of 0.8, may allow for supporting up to eighteen crew members. It has demonstrated robust fault tolerance and sustained operational efficiency, even for a 20% robot failure rate. The framework may help to address key challenges in lunar resource extraction, particularly in the permanently shadowed regions. To refine the proposed strategies, it is recommended that further studies be conducted on their large-scale applications in space mining operations at the Extraterrestrial Environmental Simulation (EXTERRES) laboratory at the University of Adelaide. Full article