Search Results (77)

The rapid and accurate prediction of crop yield and the construction of optimal yield prediction models are important for guiding field-scale agronomic management practices in precision agriculture. This study selected the leaf area index (LAI) of winter wheat (Triticum aestivum L.) at four different stages, and collected canopy spectral information and extracted vegetation indexes through unmanned aerial vehicle (UAV) multi-spectral sensors to establish the yield prediction model under the condition of slow-release nitrogen fertilizer and proposed optimal fertilization strategies for sustainable yield increase in wheat. The prediction results were evaluated using random forest (RF), support vector machine (SVM) and back propagation neural network (BPNN) methods to select the optimal spectral index and establish yield prediction models. The results showed that LAI has a significantly positive correlation with yield across four growth stages of winter wheat, and the correlation coefficient at the anthesis stage reached 0.96 in 2018–2019 and 0.83 in 2019–2020. Therefore, yield prediction for winter wheat could be achieved through a remote sensing estimation of LAI at the anthesis stage. Six vegetation indexes calculated from UAV-derived reflectance data were modeled against LAI, demonstrating that the red-edge vegetation index (CI_{red edge}) achieved superior accuracy in estimating LAI for winter wheat yield prediction. RF, SVM and BPNN models were used to evaluate the accuracy and precision of CI_{red edge} in predicting yield, respectively. It was found that RF outperformed both SVM and BPNN in predicting yield accuracy. The CI_{red edge} of the anthesis stage was the best vegetation index and stage for estimating yield of winter wheat based on UAV remote sensing. Under different N application rates, both predicted and measured yields exhibited a consistent trend that followed the order of SRF (slow-release N fertilizer) > SRFU1 (mixed TU and SRF at a ratio of 2:8) > SRFU2 (mixed TU and SRF at a ratio of 3:7) > TU (traditional urea). The optimum N fertilizer rate and N fertilizer type for winter wheat in this study were 220 kg ha⁻¹ and SRF, respectively. The results of this study will provide significant technical support for regional crop growth monitoring and yield prediction. Full article

FWI is a nonlinear optimization problem; significant discrepancies between the initial and true velocity models can lead to solutions converging to local optima. To address this issue, we proposed a PIRNN-based FWI method with first-arrival time constraints. Physics-informed recurrent neural networks (PIRNNs) integrate the physical processes of seismic wave propagation into recurrent neural networks, offering a novel approach for full-waveform inversion (FWI). First, the physical processes of seismic wave propagation were embedded into the recurrent neural network, enabling finite-difference solutions of the wave equation through forward propagation. Second, first-arrival time differences between synthetic and observed records were calculated, which then guided the selection of appropriate seismic traces for FWI loss computation. Additionally, the spatiotemporal gradient information recorded during the forward propagation of the recurrent neural network was utilized for backpropagation, enabling nonlinear optimization of FWI. This method avoids the local optima caused by waveform mismatches between the observed and synthetic records resulting from inaccurate initial velocity models. Numerical experiments on the BP and Marmousi velocity models demonstrated that the proposed method accurately reconstructed subsurface velocity structures even when the initial model significantly deviated from the true model, and maintained a degree of reconstruction accuracy in the presence of considerable noise, thereby validating its low sensitivity to the initial model and its robustness against noise. Full article

Leaf area index (LAI) serves as a critical indicator for evaluating crop growth and guiding field management practices. While spectral information (vegetation indices and texture features) extracted from multispectral sensors mounted on unmanned aerial vehicles (UAVs) holds promise for LAI estimation, the limitations of single-texture features necessitate further exploration. Therefore, this study conducted field experiments over two consecutive years (2021–2022) to collect winter oilseed rape LAI ground truth data and corresponding UAV multispectral imagery. Vegetation indices were constructed, and canopy texture features were extracted. Subsequently, a correlation matrix method was employed to establish novel randomized combinations of three-dimensional texture indices. By analyzing the correlations between these parameters and winter oilseed rape LAI, variables with significant correlations (p < 0.05) were selected as model inputs. These variables were then partitioned into distinct combinations and input into three machine learning models—Support Vector Machine (SVM), Backpropagation Neural Network (BPNN), and Extreme Gradient Boosting (XGBoost)—to estimate winter oilseed rape LAI. The results demonstrated that the majority of vegetation indices and texture features exhibited significant correlations with LAI (p < 0.05). All randomized texture index combinations also showed strong correlations with LAI (p < 0.05). Notably, the three-dimensional texture index NDTTI exhibited the highest correlation with LAI (R = 0.725), derived from the spatial combination of DIS5, VAR5, and VAR3. Integrating vegetation indices, texture features, and three-dimensional texture indices as inputs into the XGBoost model yielded the highest estimation accuracy. The validation set achieved a determination coefficient (R²) of 0.882, a root mean square error (RMSE) of 0.204 cm²cm⁻², and a mean relative error (MRE) of 6.498%. This study provides an effective methodology for UAV-based multispectral monitoring of winter oilseed rape LAI and offers scientific and technical support for precision agriculture management practices. Full article

In this work, we explore an optimization idea for the complexity guidance of a phase retrieval solution for a single acquired hologram. This method associates free-space backpropagation with the fast iterative shrinkage-thresholding algorithm (FISTA), which incorporates an improvement in the total variation (TV) to guide the complexity of the phase retrieval solution from the complex diffracted field measurement. The developed procedure can provide excellent phase reconstruction using only a single acquired hologram. Full article

This study proposes a three-dimensional (3D) residual attention network (3DRFNet) for human activity recognition by learning spatiotemporal representations from motion pictures. Core innovation integrates the attention mechanism into the 3D ResNet framework to emphasize key features and suppress irrelevant ones. In each 3D ResNet block, channel and spatial attention mechanisms generate attention maps for tensor segments, which are then multiplied by the input feature mapping to emphasize key features. Additionally, the integration of Fast Fourier Convolution (FFC) enhances the network’s capability to effectively capture temporal and spatial features. Simultaneously, we used the cross-entropy loss function to describe the difference between the predicted value and GT to guide the model’s backpropagation. Subsequent experimental results have demonstrated that 3DRFNet achieved SOTA performance in human action recognition. 3DRFNet achieved accuracies of 91.7% and 98.7% on the HMDB-51 and UCF-101 datasets, respectively, which highlighted 3DRFNet’s advantages in recognition accuracy and robustness, particularly in effectively capturing key behavioral features in videos using both attention mechanisms. Full article

This study investigates the classification of pigment-manufacturing processes using deep learning to identify the optimal model for cultural property preservation science. Four convolutional neural networks (CNNs) (i.e., AlexNet, GoogLeNet, ResNet, and VGG) and one vision transformer (ViT) were compared on micrograph datasets of various pigments. Classification performance indicators, receiver-operating characteristic curves, precision–recall curves, and interpretability served as the primary evaluation measures. The CNNs achieved accuracies of 97–99%, while the ViT reached 100%, emerging as the best-performing model. These findings indicate that the ViT has potential for recognizing complex patterns and correctly processing data. However, interpretability using guided backpropagation approaches revealed limitations in the ViT ability to generate class activation maps, making it challenging to understand its internal behavior through this technique. Conversely, CNNs provided more detailed interpretations, offering valuable insights into the learned feature maps and hierarchical data processing. Despite its interpretability challenges, the ViT outperformed the CNNs across all evaluation metrics. This study underscores the potential of deep learning in classifying pigment manufacturing processes and contributes to cultural property conservation science by strengthening its scientific foundation for the conservation and restoration of historical artifacts. Full article

Installing offshore wind jackets faces increasing risks from dynamic marine conditions and is challenged by trajectory deviations due to coupled hydrodynamic and environmental factors. To address the limitations of software, such as long simulation times and tedious parameter adjustments, this study develops a rapid prediction model combining Radial Basis Function (RBF) and Backpropagation (BP) neural networks. The model is enhanced by incorporating both numerical simulation data and real-world measurement data from the launching operation. The real-world data, including the barge attitude before launching, jacket weight distribution, and actual environmental conditions, are used to refine the model and guide the development of a fully parameterized adaptive controller. This controller adjusts in real time, with its performance validated against simulation results. A case study from the Pearl River Mouth Basin was conducted, where datasets—capturing termination time, six-degrees-of-freedom motion data for the barge and jacket, and actual environmental conditions—were collected and integrated into the RBF and BP models. Numerical models also revealed that wind and wave conditions significantly affected lateral displacement and rollover risks, with certain directions leading to heightened operational challenges. On the other hand, operations under more stable environmental conditions were found to be safer, although precautions were still necessary under strong environmental loads to prevent collisions between the jacket and the barge. This approach successfully reduces weather-dependent operational delays and structural load peaks. Hydrodynamic analysis highlights the importance of directional strategies in minimizing environmental impacts. The model’s efficiency, requiring a fraction of the time compared to traditional methods, makes it suitable for real-time applications. Overall, this method provides a scalable solution to enhance the resilience of marine operations in renewable energy projects, offering both computational efficiency and high predictive accuracy. Full article

The precise prediction of natural gas pipeline corrosion rates holds great significance for pipeline maintenance and corrosion control. Existing prediction methods, especially traditional models, often fail to adequately consider noise interference and the strong nonlinear characteristics of corrosion data, resulting in insufficient prediction accuracy. To enhance predictive performance, a hybrid prediction model based on machine learning is been proposed. This model consists of three main components: data processing, model optimization, and prediction performance evaluation. In this model, data decomposition algorithms and principal component analysis are employed to eliminate redundant noise from the original data and capture their primary features. A stratified sampling method is utilized to divide the data into a training set and test set, avoiding biases caused by random sampling. A modified particle swarm optimization algorithm is applied to optimize the parameters of a back propagation neural network. The model’s predictive performance is assessed using various indicators, including R², MAPE, RMSE, MAE, U1, U2, RE, forecasting effectiveness, comparing the results with existing literature, Grey Relational Analysis, and interpretability research. The proposed prediction model is compared with eight advanced prediction models using data from a natural gas pipeline in western China. This study reveals that the developed model outperforms the others, demonstrating excellent prediction accuracy and effectively guiding the formulation of corrosion control measures. Full article

An accurate state of health (SOH) assessment of lithium-ion batteries is essential for ensuring the reliability and safety of electric vehicles (EVs). Data-driven SOH estimation methods have shown promise but face challenges in generalizing across diverse battery types and variable operating conditions. To address this, this study integrates physical information into data-driven approaches, enabling physically consistent inferences and a rapid adaptation to different battery chemistries and usage scenarios. Specifically, physical features correlated with battery degradation, such as the link between incremental capacity (IC) peaks and SOH, are used as constraints to guide model learning. A fully connected layer within a back-propagation neural network (BPNN) is employed to capture battery aging dynamics effectively. Experimental results on two datasets show that the proposed model outperforms traditional neural networks, reducing the RMSE by at least 1.1% and demonstrating strong generalizability in both single-dataset and transfer learning tasks. Full article

In real-world scenarios, it is common to apply a damping layer of a specific thickness to the surface of an acoustic black hole (ABH) beam to boost its energy dissipation capacity. However, it has become apparent that excessive damping layers might result in negative consequences. The present study suggests employing the backpropagation (BP) algorithm to refine the positioning, thickness, and contour of the damping layer for optimal results. This study begins with the derivation of a semi-analytical solution for the vibration characteristics of an ABH beam under a harmonic load using the Gaussian expansion method (GEM). This process results in the preliminary identification of a thickness profile for the damping layer that exhibits significant potential for energy dissipation. Subsequently, a BP neural network is trained on the data produced by the semi-analytical solution to further optimize this thickness variation function. The findings reveal that the geometry of the damping layer has a more complex influence on performance than previously recognized. The optimization guided by the BP neural network suggests that achieving a strong ABH effect does not require uniform application of the damping layer across the entire ABH section. Rather, the most effective approach is to concentrate the damping layer thickness at the ABH tip, with a rapid decrease in thickness as one moves away from this point. It is also determined that applying a damping layer in areas far from the tip is unnecessary. Additionally, an innovative strategy is proposed to enhance the system’s energy dissipation capabilities without changing the truncation thickness of the ABH beam. This research contributes to a deeper understanding of how the damping layer affects the energy dissipation performance of ABH beams. Full article

Domain-generalizable re-identification (DG Re-ID) aims to train a model on one or more source domains and evaluate its performance on unseen target domains, a task that has attracted growing attention due to its practical relevance. While numerous methods have been proposed, most rely on discriminative or contrastive learning frameworks to learn generalizable feature representations. However, these approaches often fail to mitigate shortcut learning, leading to suboptimal performance. In this work, we propose a novel method called diffusion model-assisted representation learning with a correlation-aware conditioning scheme (DCAC) to enhance DG Re-ID. Our method integrates a discriminative and contrastive Re-ID model with a pre-trained diffusion model through a correlation-aware conditioning scheme. By incorporating ID classification probabilities generated from the Re-ID model with a set of learnable ID-wise prompts, the conditioning scheme injects dark knowledge that captures ID correlations to guide the diffusion process. Simultaneously, feedback from the diffusion model is back-propagated through the conditioning scheme to the Re-ID model, effectively improving the generalization capability of Re-ID features. Extensive experiments on both single-source and multi-source DG Re-ID tasks demonstrate that our method achieves state-of-the-art performance. Comprehensive ablation studies further validate the effectiveness of the proposed approach, providing insights into its robustness. Full article

The interpretability of neural networks has garnered significant attention. In the domain of computer vision, gradient-based feature attribution techniques like RectGrad have been proposed to utilize saliency maps to demonstrate feature contributions to predictions. Despite advancements, RectGrad falls short in category discrimination, producing similar saliency maps across categories. This paper pinpoints the ineffectiveness of threshold-based strategies in RectGrad for distinguishing feature gradients and introduces Class activation map Guided BackPropagation (CGBP) to tackle the issue. CGBP leverages class activation maps during backpropagation to enhance gradient selection, achieving consistent improvements across four models (VGG16, VGG19, ResNet50, and ResNet101) on ImageNet’s validation set. Notably, on VGG16, CGBP improves SIC, AIC, and IS scores by 10.3%, 11.5%, and 4.5%, respectively, compared to RectGrad while maintaining competitive DS performance. Moreover, CGBP demonstrates greater sensitivity to model parameter changes than RectGrad, as confirmed by a sanity check. The proposed method has broad applicability in scenarios like model debugging, where it identifies causes of misclassification, and medical image diagnosis, where it enhances user trust by aligning visual explanations with clinical insights. Full article

Currently, research on peg-in-hole (PiH) compliant assembly is predominantly limited to circular pegs and holes, with insufficient exploration of various complex-shaped PiH tasks. Furthermore, the degree of freedom for rotation about the axis of the circular peg cannot be constrained after assembly, and few studies have covered the complete process from autonomous hole-searching to insertion. Based on the above problems, a novel cross-shaped peg and hole design has been devised. The center coordinates of the cross-hole are obtained during the hole-searching process using the three-dimensional reconstruction theory of a binocular stereo vision camera. During the insertion process, 26 contact states of the cross-peg and the cross-hole were classified, and the mapping relationship between the force-moment sensor and relative errors was established based on a backpropagation (BP) neural network, thus completing the task of autonomous PiH assembly. This system avoids hand-guiding, completely realizes the autonomous assembly task from hole-searching to insertion, and can be replaced by other structures of pegs and holes for repeated assembly after obtaining the accurate relative pose between two assembly platforms, which provides a brand-new and unified solution for complex-shaped PiH assembly. Full article

Two-dimensional periodic dielectric bars have potential applications in high-power microwave (HPM) radiation effect experiments performed in wind tunnels. Such a bar is designed to consist of two types of dielectric materials, and two lined-up blocks can be considered as a period along the bar. Under plane excitation, the theoretical period length of the beat wave pattern fits well with the simulation result, which requires modifying the previously presented field-matching method. The phase distribution on the cross-section can be non-uniform when two different guiding modes are excited independently and propagate along different materials. Directional reflection with a low reflection angle can be obtained by reasonably choosing the parameters of the dielectric array. The designed array can decrease the returned-back microwave power toward the microwave source by 6 dB according to the numerical simulation, which included the wind tunnel, the input antenna, the test target, and the reflect array in one model. Full article

Biofouling is a serious problem in marine aquaculture facilities, exerting several negative effects on cage structures. In this study, different materials of nets were placed in the Fujian Sea area of China, and the main biological adhesion species were determined. The drag force of different materials of fouled nets was studied by a physical test in a flume tank. The drag force coefficient of a clean polyethylene terephthalate (PET) net was 0.53. The drag force coefficients of ultrahigh-molecular-weight polyethylene (UHMWPE) and polyethylene (PE) nets were 161.2% and 133.5% higher, respectively, compared with those of PET nets. Crustaceans, mollusks, and algae were the main organisms that adhered to the nets. Compared with the clean nets, the drag force of PET, UHMWPE, and PE nets increased by 1.29–5.06 times, 1.11–2.85 times, and 0.55–2.46 times, respectively. Based on backpropagation (BP) neural network training, the relationship between biological characteristics (average adhesion thickness and density) and the drag force of three kinds of net materials was determined. The drag force of the biofouled net at various time points throughout the year can be predicted based on this model, which can guide the cleaning and maintenance of nets in cage structures. Full article