Search Results (294)

In the era of smart agriculture, accurate identification of plant species is critical for effective crop management, biodiversity monitoring, and the sustainable use of medicinal resources. However, existing deep learning approaches often underperform when applied to fine-grained plant classification tasks due to the lack of domain-specific, high-quality datasets and the limited representational capacity of traditional architectures. This study addresses these challenges by introducing a novel, well-curated leaf image dataset consisting of 39 classes of medicinal and aromatic plants collected from the Aromatic and Medicinal Plant Research Station in Odakkali, Kerala, India. To overcome performance bottlenecks observed with a baseline Convolutional Neural Network (CNN) that achieved only 44.94% accuracy, we progressively enhanced model performance through a series of architectural innovations. These included the use of a pre-trained VGG16 network, data augmentation techniques, and fine-tuning of deeper convolutional layers, followed by the integration of Squeeze-and-Excitation (SE) attention blocks. Ultimately, we propose a hybrid deep learning architecture that combines VGG16 with Batch Normalization, Gated Recurrent Units (GRUs), Transformer modules, and Dilated Convolutions. This final model achieved a peak validation accuracy of 95.24%, significantly outperforming several baseline models, such as custom CNN (44.94%), VGG-19 (59.49%), VGG-16 before augmentation (71.52%), Xception (85.44%), Inception v3 (87.97%), VGG-16 after data augumentation (89.24%), VGG-16 after fine-tuning (90.51%), MobileNetV2 (93.67), and VGG16 with SE block (94.94%). These results demonstrate superior capability in capturing both local textures and global morphological features. The proposed solution not only advances the state of the art in plant classification but also contributes a valuable dataset to the research community. Its real-world applicability spans field-based plant identification, biodiversity conservation, and precision agriculture, offering a scalable tool for automated plant recognition in complex ecological and agricultural environments. Full article

Geographic object-based image analysis (GEOBIA) has demonstrated substantial utility in remote sensing tasks. However, its integration with deep learning remains largely confined to image-level classification. This is primarily due to the irregular shapes and fragmented boundaries of segmented objects, which limit its applicability in semantic segmentation. While convolutional neural networks (CNNs) excel at local feature extraction, they inherently struggle to capture long-range dependencies. In contrast, Transformer-based models are well suited for global context modeling but often lack fine-grained local detail. To overcome these limitations, we propose GOFENet (Geo-Object Feature Enhanced Network)—a hybrid semantic segmentation architecture that effectively fuses object-level priors into deep feature representations. GOFENet employs a dual-encoder design combining CNN and Swin Transformer architectures, enabling multi-scale feature fusion through skip connections to preserve both local and global semantics. An auxiliary branch incorporating cascaded atrous convolutions is introduced to inject information of segmented objects into the learning process. Furthermore, we develop a cross-channel selection module (CSM) for refined channel-wise attention, a feature enhancement module (FEM) to merge global and local representations, and a shallow–deep feature fusion module (SDFM) to integrate pixel- and object-level cues across scales. Experimental results on the GID and LoveDA datasets demonstrate that GOFENet achieves superior segmentation performance, with 66.02% mIoU and 51.92% mIoU, respectively. The model exhibits strong capability in delineating large-scale land cover features, producing sharper object boundaries and reducing classification noise, while preserving the integrity and discriminability of land cover categories. Full article

Detecting manipulated document images is essential for verifying the authenticity of official records and preventing document forgery. However, forgery artifacts are often subtle and localized in fine-grained regions, such as text boundaries or character outlines, where visual symmetry and structural regularity are typically expected. These manipulations can disrupt the inherent symmetry of document layouts, making the detection of such inconsistencies crucial for forgery identification. Conventional CNN-based models face limitations in capturing such edge-level asymmetric features, as edge-related information tends to weaken through repeated convolution and pooling operations. To address this issue, this study proposes an edge-focused method composed of two components: the Edge Attention (EA) layer and the Edge Concatenation (EC) layer. The EA layer dynamically identifies channels that are highly responsive to edge features in the input feature map and applies learnable weights to emphasize them, enhancing the representation of boundary-related information, thereby emphasizing structurally significant boundaries. Subsequently, the EC layer extracts edge maps from the input image using the Sobel filter and concatenates them with the original feature maps along the channel dimension, allowing the model to explicitly incorporate edge information. To evaluate the effectiveness and compatibility of the proposed method, it was initially applied to a simple CNN architecture to isolate its impact. Subsequently, it was integrated into various widely used models, including DenseNet121, ResNet50, Vision Transformer (ViT), and a CAE-SVM-based document forgery detection model. Experiments were conducted on the DocTamper, Receipt, and MIDV-2020 datasets to assess classification accuracy and F1-score using both original and forged text images. Across all model architectures and datasets, the proposed EA–EC method consistently improved model performance, particularly by increasing sensitivity to asymmetric manipulations around text boundaries. These results demonstrate that the proposed edge-focused approach is not only effective but also highly adaptable, serving as a lightweight and modular extension that can be easily incorporated into existing deep learning-based document forgery detection frameworks. By reinforcing attention to structural inconsistencies often missed by standard convolutional networks, the proposed method provides a practical solution for enhancing the robustness and generalizability of forgery detection systems. Full article

Ensuring the structural safety of steel trusses in escalators is critical for the reliable operation of vertical transportation systems. While manual inspection remains widely used, its dependence on human judgment leads to extended cycle times and variable defect-recognition rates, making it less reliable for identifying subtle surface imperfections. To address these limitations, a novel context-aware, multi-scale deep learning framework based on the YOLOv5 architecture is proposed, which is specifically designed for automated structural defect detection in escalator steel trusses. Firstly, a method called GIES is proposed to synthesize pseudo-multi-channel representations from single-channel grayscale images, which enhances the network’s channel-wise representation and mitigates issues arising from image noise and defocused blur. To further improve detection performance, a context enhancement pipeline is developed, consisting of a local feature module (LFM) for capturing fine-grained surface details and a global context module (GCM) for modeling large-scale structural deformations. In addition, a multi-scale feature fusion module (MSFM) is employed to effectively integrate spatial features across various resolutions, enabling the detection of defects with diverse sizes and complexities. Comprehensive testing on the NEU-DET and GC10-DET datasets reveals that the proposed method achieves 79.8% mAP on NEU-DET and 68.1% mAP on GC10-DET, outperforming the baseline YOLOv5s by 8.0% and 2.7%, respectively. Although challenges remain in identifying extremely fine defects such as crazing, the proposed approach offers improved accuracy while maintaining real-time inference speed. These results indicate the potential of the method for intelligent visual inspection in structural health monitoring and industrial safety applications. Full article

This research presents DuSAFNet, a lightweight deep neural network for fine-grained bird audio classification. DuSAFNet combines dual-path feature fusion, spectral–temporal attention, and a multi-band ArcMarginProduct classifier to enhance inter-class separability and capture both local and global spectro–temporal cues. Unlike single-feature approaches, DuSAFNet captures both local spectral textures and long-range temporal dependencies in Mel-spectrogram inputs and explicitly enhances inter-class separability across low, mid, and high frequency bands. On a curated dataset of 17,653 three-second recordings spanning 18 species, DuSAFNet achieves 96.88% accuracy and a 96.83% F1 score using only 6.77 M parameters and 2.275 GFLOPs. Cross-dataset evaluation on Birdsdata yields 93.74% accuracy, demonstrating robust generalization to new recording conditions. Its lightweight design and high performance make DuSAFNet well-suited for edge-device deployment and real-time alerts for rare or threatened species. This work lays the foundation for scalable, automated acoustic monitoring to inform biodiversity assessments and conservation planning. Full article

Hyperspectral image change detection (HSI-CD) provides substantial value in environmental monitoring, urban planning and other fields. In recent years, deep-learning based HSI-CD methods have made remarkable progress due to their powerful nonlinear feature learning capabilities, yet they face several challenges: mixed-pixel phenomenon affecting pixel-level detection accuracy; heterogeneous spatial scales of change targets where coarse-grained features fail to preserve fine-grained details; and dependence on high-quality labels. To address these challenges, this paper introduces WSCDNet, a weakly supervised HSI-CD network employing coarse-to-fine feature learning, with key innovations including: (1) A dual-branch detection framework integrating binary and multiclass change detection at the sub-pixel level that enhances collaborative optimization through a cross-feature coupling module; (2) introduction of multi-granularity aggregation and difference feature enhancement module for detecting easily confused regions, which effectively improves the model’s detection accuracy; and (3) proposal of a weakly supervised learning strategy, reducing model sensitivity to noisy pseudo-labels through decision-level consistency measurement and sample filtering mechanisms. Experimental results demonstrate that WSCDNet effectively enhances the accuracy and robustness of HSI-CD tasks, exhibiting superior performance under complex scenarios and weakly supervised conditions. Full article

The rapid integration of complex sensors and electronic control units (ECUs) in autonomous vehicles significantly increases cybersecurity risks in vehicular networks. Although the Controller Area Network (CAN) is efficient, it lacks inherent security mechanisms and is vulnerable to various network attacks. The traditional Intrusion Detection System (IDS) makes it difficult to effectively deal with the dynamics and complexity of emerging threats. To solve these problems, a lightweight vehicular network intrusion detection framework based on Dynamic Feature Fusion Federated Learning (DFF-FL) is proposed. The proposed framework employs a two-stream architecture, including a transformer-augmented autoencoder for abstract feature extraction and a lightweight CNN-LSTM–Attention model for preserving temporal and local patterns. Compared with the traditional theoretical framework of the federated learning, DFF-FL first dynamically fuses the deep feature representation of each node through the transformer attention module to realize the fine-grained cross-node feature interaction in a heterogeneous data environment, thereby eliminating the performance degradation caused by the difference in feature distribution. Secondly, based on the final loss $L_{AE}\left(X,\widehat{X}\right)$ index of each node, an adaptive weight adjustment mechanism is used to make the nodes with excellent performance dominate the global model update, which significantly improves robustness against complex attacks. Experimental evaluation on the CAN-Hacking dataset shows that the proposed intrusion detection system achieves more than 99% F1 score with only 1.11 MB of memory and 81,863 trainable parameters, while maintaining low computational overheads and ensuring data privacy, which is very suitable for edge device deployment. Full article

The salinization of agricultural soils is a serious threat to farming and ecological balance in arid and semi-arid regions. Accurate estimation of soil water-soluble ions (calcium, carbonate, magnesium, and sulfate) is necessary for correct monitoring of soil salinization and sustainable land management. Hyperspectral ground-based data are valuable in soil salinization monitoring, but the acquisition cost is high, and the coverage is small. Therefore, this study proposes a two-stage deep learning framework with multispectral remote-sensing images. First, the wavelet transform is used to enhance the Transformer and extract fine-grained spectral features to reconstruct the ground-based hyperspectral data. A comparison of ground-based hyperspectral data shows that the reconstructed spectra match the measured data in the 450–998 nm range, with R² up to 0.98 and MSE = 0.31. This high similarity compensates for the low spectral resolution and weak feature expression of multispectral remote-sensing data. Subsequently, this enhanced spectral information was integrated and fed into a novel multiscale self-attentive Transformer model (MSATransformer) to invert four water-soluble ions. Compared with BPANN, MLP, and the standard Transformer model, our model remains robust across different spectra, achieving an R² of up to 0.95 and reducing the average relative error by more than 30%. Among them, for the strongly responsive ions magnesium and sulfate, R² reaches 0.92 and 0.95 (with RMSE of 0.13 and 0.29 g/kg, respectively). For the weakly responsive ions calcium and carbonate, R² stays above 0.80 (RMSE is below 0.40 g/kg). The MSATransformer framework provides a low-cost and high-accuracy solution to monitor soil salinization at large scales and supports precision farmland management. Full article

Crop disease identification is a pivotal research area in smart agriculture, forming the foundation for disease mapping and targeted prevention strategies. Among the most prevalent global wheat diseases, powdery mildew—caused by fungal infection—poses a significant threat to crop yield and quality, making early and accurate detection crucial for effective management. In this study, we present QY-SE-MResNet34, a deep learning-based classification model that builds upon ResNet34 to perform multi-class classification of wheat leaf images and assess powdery mildew severity at the single-leaf level. The proposed methodology begins with dataset construction following the GBT 17980.22-2000 national standard for powdery mildew severity grading, resulting in a curated collection of 4248 wheat leaf images at the grain-filling stage across six severity levels. To enhance model performance, we integrated transfer learning with ResNet34, leveraging pretrained weights to improve feature extraction and accelerate convergence. Further refinements included embedding a Squeeze-and-Excitation (SE) block to strengthen feature representation while maintaining computational efficiency. The model architecture was also optimized by modifying the first convolutional layer (conv1)—replacing the original 7 × 7 kernel with a 3 × 3 kernel, adjusting the stride to 1, and setting padding to 1—to better capture fine-grained leaf textures and edge features. Subsequently, the optimal training strategy was determined through hyperparameter tuning experiments, and GrabCut-based background processing along with data augmentation were introduced to enhance model robustness. In addition, interpretability techniques such as channel masking and Grad-CAM were employed to visualize the model’s decision-making process. Experimental validation demonstrated that QY-SE-MResNet34 achieved an 89% classification accuracy, outperforming established models such as ResNet50, VGG16, and MobileNetV2 and surpassing the original ResNet34 by 11%. This study delivers a high-performance solution for single-leaf wheat powdery mildew severity assessment, offering practical value for intelligent disease monitoring and early warning systems in precision agriculture. Full article

Hyperspectral sparse unmixing (SU) networks have recently received considerable attention due to their model hyperspectral images (HSIs) with a priori spectral libraries and to capture nonlinear features through deep networks. This method effectively avoids errors associated with endmember extraction, and enhances the unmixing performance via nonlinear modeling. However, two major challenges remain: the use of large spectral libraries with high coherence leads to computational redundancy and performance degradation; moreover, certain feature extraction models, such as Transformer, while exhibiting strong representational capabilities, suffer from high computational complexity. To address these limitations, this paper proposes a hyperspectral unmixing dual-branch network integrating an adaptive hop-aware GCN and neighborhood offset Mamba that is termed DGMNet. Specifically, DGMNet consists of two parallel branches. The first branch employs the adaptive hop-neighborhood-aware GCN (AHNAGC) module to model global spatial features. The second branch utilizes the neighborhood spatial offset Mamba (NSOM) module to capture fine-grained local spatial structures. Subsequently, the designed Mamba-enhanced dual-stream feature fusion (MEDFF) module fuses the global and local spatial features extracted from the two branches and performs spectral feature learning through a spectral attention mechanism. Moreover, DGMNet innovatively incorporates a spectral-library-pruning mechanism into the SU network and designs a new pruning strategy that accounts for the contribution of small-target endmembers, thereby enabling the dynamic selection of valid endmembers and reducing the computational redundancy. Finally, an improved ESS-Loss is proposed, which combines an enhanced total variation (ETV) with an $l_{1/2}$ sparsity constraint to effectively refine the model performance. The experimental results on two synthetic and five real datasets demonstrate the effectiveness and superiority of the proposed method compared with the state-of-the-art methods. Notably, experiments on the Shahu dataset from the Gaofen-5 satellite further demonstrated DGMNet’s robustness and generalization. Full article

Hyperspectral image (HSI) classification faces challenges such as high dimensionality, spectral redundancy, and difficulty in modeling the coupling between spectral and spatial features. Existing methods fail to fully exploit first-order derivatives and frequency domain information, which limits classification performance. To address these issues, this paper proposes a Differential-Frequency Attention-based Band Selection Transformer (DFAST) for HSI classification. Specifically, a Differential-Frequency Attention-based Band Selection Embedding Module (DFASEmbeddings) is designed to extract original spectral, first-order derivative, and frequency domain features via a multi-branch structure. Learnable band selection attention weights are introduced to adaptively select important bands, capture critical spectral information, and significantly reduce redundancy. A 3D convolution and a spectral–spatial attention mechanism are applied to perform fine-grained modeling of spectral and spatial features, further enhancing the global dependency capture of spectral–spatial features. The embedded features are then input into a cascaded Transformer encoder (SCEncoder) for deep modeling of spectral–spatial coupling characteristics to achieve classification. Additionally, learnable attention weights for band selection are outputted for dimensionality reduction. Experiments on several public hyperspectral datasets demonstrate that the proposed method outperforms existing CNN and Transformer-based approaches in classification performance. Full article

Deep learning approaches for pork freshness grading typically require large datasets, which limits their practical application due to the high costs associated with data collection. To address this challenge, we propose BBSNet, a lightweight few-shot learning model designed for accurate freshness classification with a limited number of images. BBSNet incorporates a batch channel normalization (BCN) layer to enhance feature distinguishability and employs BiFormer for optimized fine-grained feature extraction. Trained on a dataset of 600 pork images graded by microbial cell concentration, BBSNet achieved an average accuracy of 96.36% in a challenging 5-way 80-shot task. This approach significantly reduces data dependency while maintaining high accuracy, presenting a viable solution for cost-effective real-time pork quality monitoring. This work introduces a novel framework that connects laboratory freshness indicators to industrial applications in data-scarce conditions. Future research will investigate its extension to various food types and optimization for deployment on portable devices. Full article

Accurate quantification of above-ground biomass (AGB) and carbon sequestration is vital for monitoring terrestrial ecosystem dynamics, informing climate policy, and supporting carbon neutrality initiatives. However, conventional methods—ranging from manual field surveys to remote sensing techniques based solely on 2D vegetation indices—often fail to capture the intricate spectral and structural heterogeneity of forest canopies, particularly at fine spatial resolutions. To address these limitations, we introduce ForestIQNet, a novel end-to-end multimodal deep learning framework designed to estimate AGB and associated carbon stocks from UAV-acquired imagery with high spatial fidelity. ForestIQNet combines dual-stream encoders for processing multispectral UAV imagery and a voxelized Canopy Height Model (CHM), fused via a Cross-Attentional Feature Fusion (CAFF) module, enabling fine-grained interaction between spectral reflectance and 3D structure. A lightweight Transformer-based regression head then performs multitask prediction of AGB and CO₂e, capturing long-range spatial dependencies and enhancing generalization. Proposed method achieves an R² of 0.93 and RMSE of 6.1 kg for AGB prediction, compared to 0.78 R² and 11.7 kg RMSE for XGBoost and 0.73 R² and 13.2 kg RMSE for Random Forest. Despite its architectural complexity, ForestIQNet maintains a low inference cost (27 ms per patch) and generalizes well across species, terrain, and canopy structures. These results establish a new benchmark for UAV-enabled biomass estimation and provide scalable, interpretable tools for climate monitoring and forest management. Full article

Karst mountain areas, as complex geological systems formed by carbonate rock development, possess unique three-dimensional spatial structures and hydrogeological processes that fundamentally influence regional ecosystem evolution, land resource assessment, and sustainable development strategy formulation. In recent years, through the implementation of systematic ecological restoration projects, the ecological degradation of karst mountain areas in Southwest China has been significantly curbed. However, the research on the fine-grained land use mapping and quantitative characterization of spatial heterogeneity in karst mountain areas is still insufficient. This knowledge gap impedes scientific decision-making and precise policy formulation for regional ecological environment management. Hence, this paper proposes a novel methodology for land use mapping in karst mountain areas using very high resolution (VHR) remote sensing (RS) images. The innovation of this method lies in the introduction of strategies of geographical zoning and stratified object extraction. The former divides the complex mountain areas into manageable subregions to provide computational units and introduces a priori data for providing constraint boundaries, while the latter implements a processing mechanism with a deep learning (DL) of hierarchical semantic boundary-guided network (HBGNet) for different geographic objects of building, water, cropland, orchard, forest-grassland, and other land use features. Guanling and Zhenfeng counties in the Huajiang section of the Beipanjiang River Basin, China, are selected to conduct the experimental validation. The proposed method achieved notable accuracy metrics with an overall accuracy (OA) of 0.815 and a mean intersection over union (mIoU) of 0.688. Comparative analysis demonstrated the superior performance of advanced DL networks when augmented with priori knowledge in geographical zoning and stratified object extraction. The approach provides a robust mapping framework for generating fine-grained land use data in karst landscapes, which is beneficial for supporting academic research, governmental analysis, and related applications. Full article

Insect detection under real-field conditions remains a challenging task due to factors such as lighting variations and the small size of insects that often lack sufficient visual features for reliable identification by deep learning models. These limitations become especially pronounced in lightweight architectures, which, although efficient, struggle to capture fine-grained details under suboptimal conditions, such as variable lighting conditions, shadows, small object size and occlusion. To address this, we introduce the motion module, a lightweight component designed to enhance object detection by integrating motion information directly at the feature map level within the YOLOv8 backbone. Unlike methods that rely on frame differencing and require additional preprocessing steps, our approach operates on raw input and uses only two consecutive frames. Experimental evaluations demonstrate that incorporating the motion module leads to consistent performance improvements across key metrics. For instance, on the YOLOv8n model, the motion module yields gains of up to 5.11% in mAP50 and 7.83% in Recall, with only a small computational overhead. Moreover, under simulated illumination shifts using HSV transformations, our method exhibits robustness to these variations. These results highlight the potential of the motion module as a practical and effective tool for improving insect detection in dynamic and unpredictable field scenarios. Full article