Search Results (205)

High-resolution remote sensing object detection is a fundamental task in Earth observation. However, the storage, transmission, and downstream processing of high-resolution images impose substantial bandwidth, memory, and computational burdens. Compressed sensing (CS) samples and compresses the signals simultaneously, thereby reducing data transmission and storage overhead. Unfortunately, most existing CS-based pipelines require explicit image reconstruction before downstream inference, leading to heavy computational overheads and poor scalability for high-resolution remote sensing images (RSIs). This work focuses on post-acquisition image compression and explores algorithmically, rather than from a physical hardware implementation perspective, whether explicit image reconstruction is an indispensable intermediate step prior to object detection. To this end, we propose HR²SIOD-CL, an end-to-end compressed learning (CL) framework that performs object detection directly on CS measurements of high-resolution RSIs without explicit image reconstruction. HR²SIOD-CL integrates an entropy-driven content-aware adaptive sampling strategy and a measurement-domain detection backbone for multi-scale feature extraction. Although jointly optimized during training, the adaptive sampling and detection modules can be decoupled for flexible deployment. For fair comparison, an extra lightweight reconstruction network equipped with a single-step data-consistency correction is constructed as the baseline. Extensive experiments on the NWPU VHR-10 and DIOR datasets show that across various sampling ratios, HR²SIOD-CL surpasses the reconstruction-based detection method when integrated into two-stage detectors, and achieves comparable or superior detection performance to the reconstruction-based counterparts when integrated into single-stage detectors. Meanwhile, its computational overhead and GPU memory consumption are merely 6.97% and 35.74% of those of the reconstruction-based counterpart, respectively. These results indicate that CS measurements can function as an effective intermediate representation for object detection, and explicit image reconstruction is not a prerequisite when detection is the primary objective. Full article

Information analysis recommendation differs from conversational recommender systems (CRS) because relevance changes with the decision phase. The same event may support observation, interpretation, option selection, or action feedback, yet most large language model (LLM)-agent CRS represent dialogue state as intent and preference. This study examines whether explicit decision-stage awareness improves recommendation and whether it can be added independently of the LLM backbone. We propose Stage-Aware Conversational Recommender System (SA-CRS), a plug-in layer guided by the Observe–Orient–Decide–Act cycle. It decouples stage detection from LLM reasoning and uses detected stages to guide dialogue strategy and candidate re-ranking. We evaluate SA-CRS on an information analysis recommendation dataset from event-structured reports, using multi-turn simulated dialogues and four LLM backbones. Oracle stage injection improves Hit@5 by 3.0 percentage points (pp), showing that decision stage provides a signal beyond topic matching. With a prompt-based detector, SA-CRS improves Hit@5 by 9.0 pp on a strong backbone; with an independent Bidirectional Encoder Representations from Transformers (BERT) detector and probabilistic re-ranking, gains range from 6.5 to 15.5 pp. Negative controls with uniform or random stage signals fail to reproduce the improvements and may reduce efficiency. The results suggest that, within the evaluated single-domain information-analysis setting, decoupled decision-stage awareness is practical for decision-intensive CRS. Full article

LiDAR–camera fusion is widely used for 3D perception in intelligent connected vehicles, but a clean camera branch does not necessarily compensate for structured LiDAR corruption. We propose CMS-Attack, a cross-modal search framework in which only the LiDAR point cloud is perturbed while the camera input remains unchanged; here, “cross-modal” denotes that a single-modality LiDAR perturbation propagates through the LiDAR–camera fusion process and disrupts the multimodal detector, rather than simultaneous perturbation of both modalities. The framework has the following two access-dependent routes: the gray-box route contains FB-CMS, which uses camera-BEV, LiDAR-BEV, fused-BEV, and detection-head responses to construct a target-aware prior and prune an over-complete candidate pool, and Adaptive CMS, which substitutes architecture-specific intermediate responses; the decision-only black-box route contains FC-CMS, which refines candidates solely from display-level target states. On the nuScenes validation split, FB-CMS reduced matched target confidence from 0.80 to 0.03 under 140 injected points, corresponding to a 96.1% relative drop and 100% ASR@0.3. The ten-query FC-CMS achieved 58.97% ASR@0.3, compared with 6.17% for random frustum spoofing. On the query-based FUTR3D detector, Adaptive CMS reduced the mean matched-target score from 0.642 to 0.058, corresponding to a 90.97% relative reduction and 93.42% ASR@0.3. Intermediate camera-, LiDAR-, and fused-BEV region energies changed by less than 0.8% despite target suppression, indicating disruption at the fusion-decision stage (defined here as the decoder/detection-head and post-processing path from fused representations to final object predictions) rather than a collapse of BEV feature magnitude. These results show that structured LiDAR fabrication is substantially more disruptive than information removal and that generic outlier filtering incurs a robustness–accuracy tradeoff. Full article

Small-object detection in unmanned aerial vehicle (UAV) remote sensing imagery is challenged by dense target distributions, substantial scale variation, complex ground backgrounds, and limited edge-computing resources. To address these challenges, we propose CDF-DETR, an end-to-end detector derived from the Real-Time Detection Transformer (RT-DETR). First, a Cross-Stage Partial Single-Head Attention Transformer (CSP-SHAT) backbone combines efficient local feature extraction with partial-channel global interaction to improve multi-scale representation while reducing the parameter count of the backbone. Second, a dual-scale feature calibration (DSFC) module sequentially performs contextual aggregation and deformable spatial alignment, thereby improving the consistency of shallow localization features and deep semantic features. Third, Focaler-MPDIoU integrates coordinate-sensitive regression with IoU-quality-based sample reweighting for dense small-object localization. Experiments on the VisDrone-2019 test set and the UAVDT and HIT-UAV validation sets demonstrate ${\mathrm{mAP}}_{50}$ improvements of 3.1, 1.4, and 3.0 percentage points, respectively, over the RT-DETR-R18 baseline. On the VisDrone-2019 validation set, CDF-DETR improves ${\mathrm{mAP}}_{50–95}$ from 26.20% to 28.52%, corresponding to a gain of 2.32 percentage points, while reducing the parameter count by 25.7%. A compressed INT8 variant achieves 20.84 FPS for an offline image-level pipeline on an NVIDIA Jetson Orin Nano using ONNX and TensorRT. These results demonstrate improved detection accuracy with a reduced parameter footprint for UAV remote sensing image analysis. Full article

Methane (CH₄) emissions from dairy cattle contribute substantially to agricultural greenhouse gas production and are influenced by genetic, physiological, environmental and management-related factors. The present study investigated CH₄-related traits and genotype–system interactions in Holstein Friesian (HF) dairy cows using repeated laser methane detector (LMD)-based measurements. A total of 134 cows from one research herd reflecting a commercial production system were repeatedly recorded for CH₄ traits (739 observations per trait) between 2020 and 2024 and linked with milk performance test-day data, behavioral observations, environmental measurements and genomic breeding values. CH₄ traits were derived separately for respiration- and eructation-related emissions. Generalized linear mixed models revealed significant effects of wind speed, rumination behavior, interaction behavior and days in milk on several CH₄ traits. Across lactation, respiration-related CH₄ traits slightly decreased, whereas eructation-related traits increased toward later lactation stages. Correlations between CH₄-related breeding values and production traits were generally low to moderately negative, ranging from −0.24 to 0.08, indicating that selection for reduced CH₄ emissions may be achievable without major unfavorable effects on milk production traits. To evaluate the complex relationships among CH₄ emissions, production, behavior, environment, diet and genetic background, a structural equation model (SEM) was applied. Environmental conditions, particularly temperature and humidity, showed the strongest positive association with CH₄ emissions, while eructation-related CH₄ traits contributed more strongly to the latent CH₄ construct than respiration-related traits. Behavioral activity, especially rumination, indicated relevant associations with CH₄ expressions. The SEM further suggested that CH₄ emissions are shaped by interconnected environmental, physiological and genetic pathways rather than by a single dominant factor. Overall, the results highlight the importance of environmental sensitivity and longitudinal biological variation in CH₄ phenotypes under commercial dairy production conditions. Repeated on-farm CH₄ measurements, particularly eructation-associated traits, may provide valuable indicator traits for future genomic breeding and management strategies to reduce the environmental footprint of dairy cattle production. Full article

Wild arrowhead (Sagittaria trifolia L.) often occurs as isolated plants or compact clusters in paddy fields, yet these states are difficult to distinguish in unmanned aerial vehicle (UAV) imagery because they share similar color, texture, and leaf morphology. This study presents a two-stage framework in which an unchanged YOLOv8n detector localizes candidate targets and a dedicated Patch-cls network refines the single- or multiple-plant label. The classifier combines multi-level features, local multi-scale enhancement, and channel attention; a GhostConv variant is also evaluated to examine the efficiency trade-off. Annotation-box and detector-generated-box results are reported separately, followed by a complete-system evaluation that retains missed targets, false positives, duplicate detections, localization errors, and classification errors. Across three random seeds, the proposed Patch-cls obtained 94.00 ± 0.34% accuracy, 83.81 ± 1.19% Macro-F1, and 73.18 ± 3.91% multiple-class Recall. In the complete test pipeline, Macro-F1 increased from 0.5459 to 0.5539 and multiple-class F1 from 0.4224 to 0.4384, while mean average precision at an intersection over union (IoU) of 0.50 (mAP50) decreased slightly from 0.7023 to 0.7016. The optimized pipeline achieved 88.89 frames per second (FPS) on an NVIDIA RTX A4000 with approximately 1.62 GB peak allocated graphics processing unit (GPU) memory. The results indicate that Patch-cls can improve category balance under detector-generated crops, although the overall gain is modest and does not replace the need for stronger localization and dense-target separation. Full article

Object detection in UAV aerial images remains fundamentally constrained by extremely small object scales, strong background interference, and progressive structural information degradation along the feature extraction pipeline. Current small-object detection methods suffer from two fundamental deficiencies rooted in their convolutional feature extraction pipelines: the smoothing effect of strided convolutions in early layers, which attenuates fine-grained details before backbone processing, and the feature overwriting phenomenon, where sequential transformations progressively erase structural information from earlier layers. We propose PFE-Det (Progressive Feature Evolution Detector), built upon the DEIM framework and constructing a continuous optimization pathway across three stages. A Feature Adaptive Enhancement Network (FAENet) is adopted as a front-end preprocessor to decouple high- and low-frequency components via a Laplacian pyramid at the input stage, mitigating early-layer smoothing at the input. A Multi-Receptive-Field Adaptive Fusion Module (MFAM) is designed to reorganize single-path features into structure-retaining and progressive enhancement paths and is further coupled with hierarchical receptive-field modeling, suppressing feature overwriting through multi-scale context modeling. A Multi-Path Gated State Space Modeling Block (MG-SSM Block) couples HSM-SSD-based long-range dependency extraction with Convolutional Gated Linear Units (CGLU) for adaptive feature selection in the encoder. Experiments on VisDrone2019 demonstrate an AP of 0.225 and an ${\mathrm{AP}}_s$ of 0.134, yielding a 13.5% relative improvement in small-object precision over the baseline. Cross-dataset evaluations on DIOR and UAVVaste, each under independent training and testing, further support the effectiveness of progressive feature evolution for UAV small-object detection. Full article

Lightweight real-time detectors must balance long-range contextual modeling, cross-scale relational reasoning, and deployment efficiency. We present MambaACE-YOLO, an integration-oriented framework that separates computation into intra-scale encoding, cross-scale relational reasoning, sparse multi-stage feature distribution, and multi-scale detection decoding. Its primary contribution is the coordinated integration of prior frequency-decoupled and state-space ideas with detection-specific interfaces, target-scale leave-one-source aggregation, and physically prunable feature distribution, rather than a new frequency-transform, state-space, or hypergraph mechanism class. Building on prior frequency-decoupled hybrid visual Mamba research, its detection-oriented D-MobileMamba backbone applies bidirectional state-space scanning only to low-frequency components, while directional high-frequency correction and multi-kernel depthwise convolutions preserve boundaries and local shape. Compact Partial-Channel HyperACE (CP-HyperACE) models cross-scale higher-order relations in a semantic subspace and uses target-scale leave-one-source aggregation. Selective Additive FullPAD (SA-FullPAD) projects each cross-scale increment once and selects injection paths through static, physically prunable gates. On MS COCO 2017 val, the unpruned MambaACE-YOLO-N achieves 42.5 AP with 2.6 M parameters, and MambaACE-YOLO-S achieves 48.8 AP with 9.0 M parameters. Under a common documented RTX 5090 TensorRT FP16 setting, the unpruned Nano model records 1.19 ms network-forward latency at 42.5 AP, whereas its physically pruned counterpart retains 42.4 AP and records 1.02 ms. Accuracy and latency values are single-run or single-record point estimates without reported variance, and the 42.5-AP result uses the 600-epoch schedule without a matched 600-epoch YOLOv13-N control. We distinguish published cross-paper results from same-framework, same-device measurements and assess the individual design choices through controlled ablations and physical-pruning experiments. Full article

Traditional Bayesian SAR detection methods suffer poor adaptability to speckle noise, fail to handle severe class imbalance within large-scale multi-target datasets, and incur prohibitive training overheads. To address these drawbacks, this paper develops a lightweight Bayesian detection and recognition framework built upon heavy-tailed Laplacian priors and variational inference. We adopt ResNet-50 as the feature extraction backbone and design a four-stage pipeline: First, a noise-aware Laplacian heavy-tailed prior is proposed to strengthen resistance against speckle outliers. Second, a multi-class variational inference module is constructed to eliminate detection bias induced by uneven sample distribution across target categories. Third, a lightweight uncertainty feedback strategy is introduced to cut computational costs for large-batch training. Evaluated on the MSAR-1.0 dataset, our approach achieves an mAP@0.5 of 94.98% and a macro balanced accuracy ($BA$) of 93.34%. Compared with existing Bayesian detectors, the mAP metric rises by 5.44–6.53%. The model only consumes 4.33 ms per inference frame and completes full training within 1.53 h on a single GPU. Ablation tests validate the independent and combined efficacy of all three core modules. This integrated architecture balances detection precision, classification reliability, and training efficiency, offering a promising prototype for multi-class SAR target interpretation under the evaluated benchmark constraints. Full article

Manual visual diagnosis of apple diseases suffers from low efficiency, strong subjectivity and poor scalability for large commercial orchards. Existing research mainly targets diseases on single plant organs, whereas full-growth-cycle detection has to cope with extreme multi-scale differences among lesions. For instance, Valsa canker on tree trunks leads to extensive cortical necrosis, while early-stage anthracnose on fruits appears as tiny spots spanning only a few pixels. These significant scale gaps necessitate robust spatial feature aggregation and anti-noise ability to resist complex background interference. Aiming to achieve rapid and precise detection of diseases on multiple apple organs including leaves, fruits, trunks and branches, this work presents an enhanced YOLOv11 model equipped with the Information Integration Attention (IIA) module. The IIA module is embedded into the key fusion layers of the backbone and neck networks. It strengthens the extraction of fine-grained lesion features, recovers spatial location information via a bidirectional attention mechanism, and suppresses noise induced by uneven lighting and intricate backgrounds. To guarantee stable convergence on low-resource computing devices, a tailored training scheme is designed. Experimental results on a seven-category dataset with 7406 images demonstrate that YOLOv11-IIA reaches a precision of 0.763, a recall of 0.819, mAP@50 of 0.857 and mAP@50-95 of 0.699, which achieves clear performance improvements over the original YOLOv11 (mAP@50 improved from 0.485 to 0.857) and other attention-augmented detectors. The model operates stably on an NVIDIA GTX 1050 4GB GPU with an inference speed of 16 FPS for 640 × 640 input images; comprehensive quantitative computational metrics including parameter count, FLOPs and memory consumption will be fully measured in subsequent UAV deployment experiments. The proposed method provides a reliable technical reference for intelligent apple disease monitoring in smart orchard systems. Full article

Accurate sheep detection supports counting, tracking, behavior analysis, and health monitoring in intelligent livestock farming. However, occlusion, scale variation, nighttime low light, and background interference can cause missed detections and inaccurate localization. This study proposes GMS-YOLO11n, an improved YOLO11n-based detector for fixed-view sheep monitoring. A spatially gated bottleneck convolution module was introduced at key backbone downsampling stages to strengthen local structural cues, while a multi-scale attention fusion module combined deep semantic information with shallow-to-intermediate details through feature alignment, cross-scale attention, and channel-wise learnable gating. A dataset of 3531 images of Small-tailed Han sheep, containing 8167 annotated instances, was used for evaluation. Across six independent runs, GMS-YOLO11n achieved precision, recall, F1-score, mAP@0.5, and mAP@0.5:0.95 of 93.62% ± 0.12%, 91.03% ± 1.37%, 92.30% ± 0.75%, 96.53% ± 0.49%, and 76.32% ± 0.57%, respectively. Compared with the COCO-pretrained YOLO11n baseline, the corresponding improvements were 1.05, 4.38, 2.79, 2.76, and 6.92 percentage points. Subset analysis showed mAP@0.5:0.95 gains of 9.83 and 7.62 percentage points under nighttime low-light and obvious-occlusion conditions. These results demonstrate improved detection performance under the challenging visual conditions represented in the evaluated single-farm, fixed-view dataset; broader generalization requires external validation. Full article

Capacitive sensors implemented in complementary metal-oxide-semiconductor (CMOS) technology are widely used in lab-on-chip (LoC), biomedical, and microfluidic systems. While most capacitive sensor interfaces are designed for high-resolution capacitance quantification, many practical applications require only binary decisions, event detection, or state discrimination. In such scenarios, conventional readout architectures introduce unnecessary circuit complexity, power consumption, latency, and data-processing overhead. This paper presents a CMOS cross-coupled-based capacitance detector (CBCD) that directly converts the imbalance between a sensing capacitance and a reference capacitance into a digital output. By exploiting regenerative positive feedback in a dynamic latch architecture, the proposed detector integrates sensing, comparison, and digitization within a single stage, eliminating the need for analog amplification, analog-to-digital conversion, frequency-based readout, and external thresholding circuitry. Circuit-level simulations show the ability to detect extremely small capacitance differences, demonstrate robust operation across a wide range of input capacitances, and achieve negligible power consumption. Process-corner, noise, and Monte Carlo analyses further verify reliable operation in the presence of device mismatch and process variations. Owing to its compact structure, digital-native output, and energy-efficient operation, the proposed CBCD is well suited for decision-driven sensing applications, including droplet presence detection, bubble monitoring, threshold-based diagnostics, event detection, and time-of-evaporation (ToE) measurements. The proposed architecture provides a scalable and low-complexity front-end solution for next-generation CMOS-integrated sensing platforms. Full article

Software-Defined Networking (SDN) centralizes control logic, improving programmability but exposing the controller to volumetric and low-rate Distributed Denial of Service (DDoS) attacks. Entropy-based detectors often raise late alarms or require significant traffic distribution changes, while machine-learning approaches impose high training and inference overhead. To address these issues, this work proposes a Statistical Mean Renyi Entropy (SMRE)-based early-warning system that amplifies micro-level disturbances in flow randomness using a tunable sensitivity weight (μ). The formulation enhances responsiveness to entropy deviations without adding computational complexity, enabling O(n) single-pass execution per monitoring window. The method was implemented on a Mininet testbed (nine switches, 64 hosts, POX controller with the L3_learning module) with mixed benign traffic and hping3/Scapy-generated UDP and TCP flood attack traffic at intensities ranging from 10 to 75%. Experimental results demonstrate that SMRE detects early-stage attacks with 94.7–98.1% accuracy, 0.8–2.3% false positive rate, and 6.5–14 ms detection latency, outperforming Shannon and classical Renyi entropy detectors. ROC analysis (AUC ≈ 0.99) and paired t-tests (p < 0.01) confirm statistical significance. Resource profiling shows negligible CPU and memory overhead, supporting real-time deployment. By eliminating model training and ensuring robust early detection, SMRE offers a lightweight and practical detection mechanism for SDN environments, whose applicability to cloud, edge, and IoT deployments will be further substantiated through validation on real traffic traces and multi-controller architectures. Full article

Urban traffic congestion remains a major barrier to sustainable mobility, necessitating a transition from static signaling to Intelligent Traffic Light Systems (ITLS). This study presents a systematic review in computer vision, artificial intelligence, and adaptive control algorithms for urban intersection optimization. Following a methodological framework informed by PRISMA guidelines, the study examines state-of-the-art architectures, including deep learning-based perception models and reinforcement learning agents. The findings indicate that, although two-stage detectors provide benchmark accuracy for vehicle perception, single-stage models and Vision Transformers offer the high-speed processing required for real-time traffic management. In addition, deep reinforcement learning enables autonomous, lane-specific optimization that outperforms traditional actuated and fixed-time systems. The review also identifies a persistent research gap in the deployment of these computational frameworks in the resource-constrained and heterogeneous infrastructures of developing countries. For the urban context of Riobamba, Ecuador, a phased implementation strategy is proposed that balances computational demands with the city’s morphological and social characteristics. By bridging the gap between high-fidelity simulation and practical field deployment, this review provides a scalable framework for improving throughput, reducing emissions, and enhancing safety. Overall, these advances offer a promising pathway toward more resilient transportation systems in rapidly evolving Andean urban centers. Full article

Reliable stored-grain pest monitoring is essential for smart granaries, yet probe-based field images pose three coupled bottlenecks: tiny and densely distributed pests in complex backgrounds, costly bounding-box annotation, and limited edge-side computing resources. To address these bottlenecks in a targeted manner, this study proposes GrainPest-SSL, an integrated framework comprising a field dataset, a lightweight detector, and a pseudo-label purification-based semi-supervised pipeline. First, to overcome the lack of realistic training data, a GrainPest dataset with 1000 field images and 21,676 annotated pest instances is constructed using multiple self-developed monitoring probes deployed in a large wheat flat granary, capturing systematic pest-monitoring images from different in-bin locations rather than a single fixed imaging point. Second, to improve small-target detection under resource constraints, a YOLOv8n-CAEMA detector is designed with a P2 detection head and tail-inserted Coordinate Attention (CA) and Efficient Multi-scale Attention (EMA), achieving 0.840 mAP@0.5 under full supervision with only 2.932 M parameters. Third, to reduce annotation dependence without adding inference-stage complexity, an offline Teacher–Student strategy with Pseudo-Label Purification Filtering (PPLF) refines pseudo-labels using confidence, size, and aspect-ratio priors; under the 30% labeled setting, GrainPest-SSL improves mAP@0.5 from 0.738 to 0.799 and mAP@0.5:0.95 from 0.322 to 0.369 on average over three random seeds. Comparisons with representative agricultural pest detectors and semi-supervised object detection (SSOD) methods further confirm the balanced accuracy–efficiency performance of GrainPest-SSL under label-limited conditions. The deployed Student detector further achieves 13.6 FPS in FP16 mode on a Jetson Orin Nano Dev Kit under the 10 W power mode, supporting scheduled pest inspection, early infestation screening, and intelligent warning in smart granary monitoring systems. Full article