Search Results (60)

Complex marine infrared images, which suffer from background interference, blurred features, and indistinct contours, hamper detection accuracy. Meanwhile, the limited computing power, storage, and energy of maritime devices require target detection models suitable for real-time detection. To address these issues, we propose DSEE-YOLO (Dynamic Ship Edge-Enhanced YOLO), an efficient lightweight infrared ship detection algorithm. It integrates three innovative modules with pruning and self-distillation: the C3k2_MultiScaleEdgeFusion module replaces the original bottleneck with a MultiEdgeFusion structure to boost edge feature expression; the lightweight DS_ADown module uses DSConv (depthwise separable convolution) to reduce parameters while preserving feature capability; and the DyTaskHead dynamically aligns classification and localization features through task decomposition. Redundant structures are pruned via LAMP (Layer-Adaptive Sparsity for the Magnitude-Based Pruning), and performance is optimized via BCKD (Bridging Cross-Task Protocol Inconsistency for Knowledge Distillation) self-distillation, yielding a lightweight, efficient model. Experimental results show the DSEE-YOLO outperforms YOLOv11n when applied to our self-constructed IRShip dataset by reducing parameters by 42.3% and model size from 10.1 MB to 3.5 MB while increasing mAP@0.50 by 2.8%, mAP@0.50:0.95 by 3.8%, precision by 2.3%, and recall by 3.0%. These results validate its high-precision detection capability and lightweight advantages in complex infrared scenarios, offering an efficient solution for real-time maritime infrared ship monitoring. Full article

This study addresses the core challenges of precise marigold corollas recognition and efficient continuous path planning under complex natural conditions (strong illumination, occlusion, adhesion) by proposing an integrated lightweight visual recognition and real-time path planning framework. We introduce MPD-YOLO, an optimized model based on YOLOv11n, incorporating (1) a Multi-scale Information Enhancement Module (MSEE) to boost feature extraction; (2) structured pruning for significant model compression (final size: 2.1 MB, 39.6% of original); and (3) knowledge distillation to recover accuracy loss post-pruning. The resulting model achieves high precision (P: 89.8%, mAP@0.5: 95.1%) with reduced computational load (3.2 GFLOPs) while demonstrating enhanced robustness in challenging scenarios—recall significantly increased by 6.8% versus YOLOv11n. Leveraging these recognition outputs, an adaptive ant colony algorithm featuring dynamic parameter adjustment and an improved pheromone strategy reduces average path planning time to 2.2 s—a 68.6% speedup over benchmark methods. This integrated approach significantly enhances perception accuracy and operational efficiency for automated marigold harvesting in unstructured environments, providing robust technical support for continuous automated operations. Full article

The reliable operation of photovoltaic (PV) systems is often compromised by surface soiling and structural damage, which reduce energy efficiency and complicate large-scale monitoring. To address this challenge, we propose a two-tiered image-classification framework that combines Vision Transformer (ViT) models, lightweight convolutional neural networks (CNNs), and knowledge distillation (KD). In Tier 1, a DINOv2 ViT-Base model is fine-tuned to provide robust high-level categorization of solar-panel images into three classes: Normal, Soiled, and Damaged. In Tier 2, two enhanced EfficientNetB0 models are introduced: (i) a KD-based student model distilled from a DINOv2 ViT-S/14 teacher, which improves accuracy from 96.7% to 98.67% for damage classification and from 90.7% to 92.38% for soiling classification, and (ii) an EfficientNetB0 augmented with Multi-Head Self-Attention (MHSA), which achieves 98.73% accuracy for damage and 93.33% accuracy for soiling. These results demonstrate that integrating transformer-based representations with compact CNN architectures yields a scalable and efficient solution for automated monitoring of the condition of PV systems, offering high accuracy and real-time applicability in inspections on solar farms. Full article

Under large-scale farming conditions, automated sow estrus detection is crucial for improving reproductive efficiency, optimizing breeding management, and reducing labor costs. Conventional estrus detection relies heavily on human expertise, a practice that introduces subjective variability and consequently diminishes both accuracy and efficiency. Failure to identify estrus promptly and pair animals effectively lowers breeding success rates and drives up overall husbandry costs. In response to the need for the automated detection of sows’ estrus states in large-scale pig farms, this study proposes a method for detecting sows’ vulvar status and estrus based on multi-dimensional feature crossing. The method adopts a dual optimization strategy: First, the Bi-directional Feature Pyramid Network—Selective Decoding Integration (BiFPN-SDI) module performs the bidirectional, weighted fusion of the backbone’s low-level texture and high-level semantic, retaining the multi-dimensional cues most relevant to vulvar morphology and producing a scale-aligned, minimally redundant feature map. Second, by embedding a Spatially Enhanced Attention Module head (SEAM-Head) channel attention mechanism into the detection head, the model further amplifies key hyperemia-related signals, while suppressing background noise, thereby enabling cooperative and more precise bounding box localization. To adapt the model for edge computing environments, Masked Generative Distillation (MGD) knowledge distillation is introduced to compress the model while maintaining the detection speed and accuracy. Based on the bounding box of the vulvar region, the aspect ratio of the target area and the red saturation features derived from a dual-threshold method in the HSV color space are used to construct a lightweight Multilayer Perceptron (MLP) classification model for estrus state determination. The network was trained on 1400 annotated samples, which were divided into training, testing, and validation sets in an 8:1:1 ratio. On-farm evaluations in commercial pig facilities show that the proposed system attains an 85% estrus detection success rate. Following lightweight optimization, inference latency fell from 24.29 ms to 18.87 ms, and the model footprint was compressed from 32.38 MB to 3.96 MB in the same machine, while maintaining a mean Average Precision (mAP) of 0.941; the accuracy penalty from model compression was kept below 1%. Moreover, the model demonstrates robust performance under complex lighting and occlusion conditions, enabling real-time processing from vulvar localization to estrus detection, and providing an efficient and reliable technical solution for automated estrus monitoring in large-scale pig farms. Full article

As demand for the precious medicinal herb ginseng continues to grow, its importance is becoming ever more prominent. Traditional manual methods are inefficient and inconsistent. Thus, improving the accuracy and efficiency of ginseng quality testing is the central objective of this study. We collected ginseng samples and expanded the dataset through augmentation, which added noise, varied lighting, and surface defects such as red rust and insect damage, to reflect real-world conditions. Because ginseng has intricate textures, irregular shapes, and unstable lighting, we built LLT-YOLO on the YOLOv11 framework, adding a DCA module, depth-wise separable convolutions, an efficient multi-scale attention mechanism, and knowledge distillation to boost accuracy on small devices. Tests showed a precision of 90.5%, a recall of 92.3%, an mAP50 of 95.1%, and an mAP50–95 of 77.4%, gains of 3%, 2.2%, 7.8%, and 0.5% over YOLOv11 with fewer parameters and smaller size, confirming LLT-YOLO as a practical tool for appearance-based ginseng grading that can be extended to other crops. The results indicate that LLT-YOLO offers a practical tool for appearance-based ginseng quality assessment and can be extended to other crops in future work. Full article

Agricultural image processing technology plays a critical role in enabling precise disease detection, accurate yield prediction, and various smart agriculture applications. However, its practical implementation faces key challenges, including environmental interference, data scarcity and imbalance datasets, and the difficulty of deploying models on resource-constrained edge devices. This paper presents a systematic review of recent advances in addressing these challenges, with a focus on three core aspects: environmental robustness, data efficiency, and model deployment. The study identifies that attention mechanisms, Transformers, multi-scale feature fusion, and domain adaptation can enhance model robustness under complex conditions. Self-supervised learning, transfer learning, GAN-based data augmentation, SMOTE improvements, and Focal loss optimization effectively alleviate data limitations. Furthermore, model compression techniques such as pruning, quantization, and knowledge distillation facilitate efficient deployment. Future research should emphasize multi-modal fusion, causal reasoning, edge–cloud collaboration, and dedicated hardware acceleration. Integrating agricultural expertise with AI is essential for promoting large-scale adoption, as well as achieving intelligent, sustainable agricultural systems. Full article

This paper addresses the challenges of anomaly detection in industrial components by proposing a two-stage deep-learning approach combining semantic segmentation and knowledge distillation. Traditional methods, such as manual inspection and machine vision, face limitations in efficiency and accuracy when dealing with complex defects. To overcome these issues, we first introduce a small-sample semantic segmentation model based on a U-Net architecture, enhanced with an Adaptive Multi-Scale Attention Module (AMAM) and gate attention mechanisms to improve edge detection and multi-scale feature extraction. The second stage employs a knowledge distillation-based anomaly detection model, where a pre-trained teacher network (WideResNet50) extracts features, and a student network reconstructs them, with differences indicating anomalies. A Transformer-based feature aggregation module further refines the process. Experiments on the MVTec dataset demonstrate superior performance, with the segmentation model achieving 96.4% mIoU and the anomaly detection model attaining 98.3% AUC, outperforming State-of-the-Art methods. Under an extremely small-sample regime of merely 27 training images, the proposed model still attains a mIoU exceeding 94%. The two-stage approach significantly enhances detection accuracy by reducing background interference and focusing on localized defects. This work contributes to industrial quality control by improving efficiency, reducing false positives, and adapting to limited annotated data. Full article

Apple-detection performance in orchards degrades markedly under low-light conditions, where intensified noise and non-uniform exposure blur edge cues critical for precise localisation. We propose Knowledge Distillation with Geometry-Consistent Feature Alignment (KDFA), a compact end-to-end framework that couples image enhancement and detection through the following two complementary components: (i) Cross-Domain Mutual-Information-Bound Knowledge Distillation, which maximises an InfoNCE lower bound between daylight-teacher and low-light-student region embeddings; (ii) Geometry-Consistent Feature Alignment, which imposes Laplacian smoothness and bipartite graph correspondences across multiscale feature lattices. Trained on 1200 pixel-aligned bright/low-light image pairs, KDFA achieves 51.3% mean Average Precision (mAPQ [0.50:0.95]) on a challenging low-light apple-detection benchmark, setting a new state of the art by simultaneously bridging the illumination-domain gap and preserving geometric consistency. Full article

UAV small target detection in urban security, disaster monitoring, agricultural inspection, and other fields faces the challenge of increasing accuracy and real-time requirements. However, existing detection algorithms still have weak small target representation ability, extensive computational resource overhead, and poor deployment adaptability. Therefore, this paper proposes a lightweight algorithm, I-YOLOv11n, based on YOLOv11n, which is systematically improved in terms of both feature enhancement and structure compression. The RFCBAMConv module that combines deformable convolution and channel–spatial attention is designed to adjust the receptive field and strengthen the edge features dynamically. The multiscale pyramid of STCMSP context and the lightweight Transformer–DyHead hybrid detection head are designed by combining the multiscale hole feature pyramid (DFPC), which realizes the cross-scale semantic modeling and adaptive focusing of the target area. A collaborative lightweight strategy is proposed. Firstly, the semantic discrimination ability of the teacher model for small targets is transferred to guide and protect the subsequent compression process by integrating the mixed knowledge distillation of response alignment, feature imitation, and structure maintenance. Secondly, the LAMP–Taylor channel pruning mechanism is used to compress the model redundancy, mainly to protect the key channels sensitive to shallow small targets. Finally, K-means++ anchor frame optimization based on IoU distance is implemented to adapt the feature structure retained after pruning and the scale distribution of small targets of UAV. While significantly reducing the model size (parameter 3.87 M, calculation 14.7 GFLOPs), the detection accuracy of small targets is effectively maintained and improved. Experiments on VisDrone, AI-TOD, and SODA-A datasets show that the mAP@0.5 and mAP@0.5:0.95 of I-YOLOv11n are 7.1% and 4.9% higher than the benchmark model YOLOv11 n, respectively, while maintaining real-time processing capabilities, verifying its comprehensive advantages in accuracy, light weight, and deployment. Full article

In real-world applications, user demands for new functionalities and activities constantly evolve, requiring action recognition systems to incrementally incorporate new action classes without retraining from scratch. This class-incremental learning (CIL) paradigm is essential for enabling adaptive and scalable systems that can grow over time. However, Wi-Fi-based indoor action recognition under incremental learning faces two major challenges: catastrophic forgetting of previously learned knowledge and uncontrolled model expansion as new classes are added. To address these issues, we propose CAREC, a class-incremental framework that balances dynamic model expansion with efficient compression. CAREC adopts a multi-branch architecture to incorporate new classes without compromising previously learned features and leverages balanced knowledge distillation to compress the model by 80% while preserving performance. A data replay strategy retains representative samples of old classes, and a super-feature extractor enhances inter-class discrimination. Evaluated on the large-scale XRF55 dataset, CAREC reduces performance degradation by 51.82% over four incremental stages and achieves 67.84% accuracy with only 21.08 M parameters, 20% parameters compared to conventional approaches. Full article

Autonomous earthwork machinery is gaining traction as a means to boost productivity and safety on space-constrained urban sites, yet the fast-growing literature has not been fully integrated. To clarify current knowledge, we systematically searched Scopus and screened 597 records, retaining 157 peer-reviewed papers (2015–March 2025) that address autonomy, integrated control, or risk mitigation for excavators, bulldozers, and loaders. Descriptive statistics, VOSviewer mapping, and qualitative synthesis show the output rising rapidly and peaking at 30 papers in 2024, led by China, Korea, and the USA. Four tightly linked themes dominate: perception-driven machine autonomy, IoT-enabled integrated control systems, multi-sensor safety strategies, and the first demonstrations of fleet-level collaboration (e.g., coordinated excavator clusters and unmanned aerial vehicle and unmanned ground vehicle (UAV–UGV) site preparation). Advances include centimeter-scale path tracking, real-time vision-light detection and ranging (LiDAR) fusion and geofenced safety envelopes, but formal validation protocols and robust inter-machine communication remain open challenges. The review distils five research priorities, including adaptive perception and artificial intelligence (AI), digital-twin integration with building information modeling (BIM), cooperative multi-robot planning, rigorous safety assurance, and human–automation partnership that must be addressed to transform isolated prototypes into connected, self-optimizing fleets capable of delivering safer, faster, and more sustainable urban construction. Full article

Automated defect detection for printed fabric manufacturing faces critical challenges in balancing industrial-grade accuracy with real-time deployment efficiency. To address this, we propose RT-DETR-FFD, a knowledge-distilled detector optimized for printed fabric defect inspection. Firstly, the student model integrates a Fourier cross-stage mixer (FCSM). This module disentangles defect features from periodic textile backgrounds through spectral decoupling. Secondly, we introduce FuseFlow-Net to enable dynamic multi-scale interaction, thereby enhancing discriminative feature representation. Additionally, a learnable positional encoding (LPE) module transcends rigid geometric constraints, strengthening contextual awareness. Furthermore, we design a dynamic correlation-guided loss (DCGLoss) for distillation optimization. Our loss leverages masked frequency-channel alignment and cross-domain fusion mechanisms to streamline knowledge transfer. Experiments demonstrate that the distilled model achieves an mAP@0.5 of 82.1%, surpassing the baseline RT-DETR-R18 by 6.3% while reducing parameters by 11.7%. This work establishes an effective paradigm for deploying high-precision defect detectors in resource-constrained industrial scenarios, advancing real-time quality control in textile manufacturing. Full article

Maize (Zea mays L.), a vital global food crop, relies on its stomatal structure for regulating photosynthesis and responding to drought. Conventional manual stomatal detection methods are inefficient, subjective, and inadequate for high-throughput plant phenotyping research. To address this, we curated a dataset of over 1500 maize leaf epidermal stomata images and developed a novel lightweight detection model, StomaYOLO, tailored for small stomatal targets and subtle features in microscopic images. Leveraging the YOLOv11 framework, StomaYOLO integrates the Small Object Detection layer P2, the dynamic convolution module, and exploits large-scale epidermal cell features to enhance stomatal recognition through auxiliary training. Our model achieved a remarkable 91.8% mean average precision (mAP) and 98.5% precision, surpassing numerous mainstream detection models while maintaining computational efficiency. Ablation and comparative analyses demonstrated that the Small Object Detection layer, dynamic convolutional module, multi-task training, and knowledge distillation strategies substantially enhanced detection performance. Integrating all four strategies yielded a nearly 9% mAP improvement over the baseline model, with computational complexity under 8.4 GFLOPS. Our findings underscore the superior detection capabilities of StomaYOLO compared to existing methods, offering a cost-effective solution that is suitable for practical implementation. This study presents a valuable tool for maize stomatal phenotyping, supporting crop breeding and smart agriculture advancements. Full article

Frequent and complex boarding operations require an effective management process for specialized tools. Traditional manual statistical analysis exhibits low efficiency, poor accuracy, and a lack of electronic records, making it difficult to meet the demands of modern aviation manufacturing. In this study, we propose an efficient and lightweight network designed for the recognition and analysis of professional tools. We employ a combination of knowledge distillation and pruning techniques to construct a compact network optimized for the target dataset and constrained deployment resources. We introduce a self-attention mechanism (SAM) for multi-scale feature fusion within the network to enhance its feature segmentation capability on the target dataset. In addition, we integrate a large language model (LLM), enhanced by retrieval-augmented generation (RAG), to analyze tool detection results, enabling the system to rapidly provide relevant information about operational tools for management personnel and facilitating intelligent monitoring and control. Experimental results on multiple benchmark datasets and professional tool datasets validate the effectiveness of our approach, demonstrating superior performance. Full article

Ensuring that electrical workers use personal protective equipment (PPE) correctly is critical to electrical safety, but existing detection methods face significant limitations when applied in the electrical industry. This paper introduces MRC-DETR (Multi-Scale Re-calibration Detection Transformer), a novel framework for detecting Power Engineering Personal Protective Equipment (PEPPE) in complex electrical operating environments. Our method introduces two technical innovations: a Multi-Scale Enhanced Boundary Attention (MEBA) module, which significantly improves the detection of small and occluded targets through optimized feature representation, and a knowledge distillation strategy that enables efficient deployment on edge devices. We further contribute a dedicated PEPPE dataset to address the lack of domain-specific training data. Experimental results demonstrate superior performance compared to existing methods, particularly in challenging power industry scenarios. Full article