Multi-feature Fusion Research Articles

Advanced State-of-Health Estimation for Lithium-Ion Batteries Using Multi-Feature Fusion and KAN-LSTM Hybrid Model

Accurate assessment of battery State of Health (SOH) is crucial for the safe and efficient operation of electric vehicles (EVs), which play a significant role in reducing reliance on non-renewable energy sources. This study introduces a novel SOH estimation method combining Kolmogorov–Arnold Networks (KAN) and Long Short-Term Memory (LSTM) networks. The method is based on fully charged battery characteristics, extracting key parameters such as voltage, temperature, and charging data collected during cycles. Validation was conducted under a temperature range of 10 °C to 30 °C and different charge–discharge current rates. Notably, temperature variations were primarily caused by seasonal changes, enabling the experiments to more realistically simulate the battery’s performance in real-world applications. By enhancing dynamic modeling capabilities and capturing long-term temporal associations, experimental results demonstrate that the method achieves highly accurate SOH estimation under various charging conditions, with low mean absolute error (MAE) and root mean square error (RMSE) values and a coefficient of determination (R2) exceeding 97%, significantly improving prediction accuracy and efficiency.

Prediction Method for Pomegranate Chlorophyll Content Based on Multi-feature Fusion of Unmanned Aerial Vehicle

The aim of this study was to obtain RGB and multispectral images of fruit tree canopy during the flowering period of pomegranate by multispectral unmanned aerial vehicle (UAV) to quickly and accurately predict the chlorophyll content in order to improve the monitoring efficiency of the orchard. A handheld chlorophyll meter was used to obtain the actual chlorophyll values, and image processing techniques were combined to extract parameters such as color features and texture features of the RGB images as well as vegetation index of the multispectral images. A chlorophyll content prediction method based on the support vector regression model and the convolutional neural network model (CNN-Attention) combined with the attention mechanism was established. The results showed that (1) the accuracy of the prediction model was improved by the fusion of RGB image features and multispectral image vegetation index. (2) After model comparison, the CNN-Attention model built under the fused features was the best in chlorophyll content prediction with R2, RE, and RMSE of 0.9699, 0.0052, and 0.6013, respectively.This study provides a more accurate method for fruit tree chlorophyll content prediction using UAVs, which provides a practical reference for orchard management.

Automatic detection of marine oil spills from polarimetric SAR images using deep Convolutional neural network model

Oil Spills (OS) exert serious threats to marine ecosystems, hence quick and accurate monitoring of OS is of great significance. In recent years, the Synthetic Aperture Radar (SAR) remote sensing technique has been used to monitor OS because of its wide coverage and operating in any weather conditions. However, many studies only resort to manual interpretation when morning OS by SAR images, making the interpretation process low efficiency. In this study, an intelligent method is presented for detecting OS from dual-polarized SAR images using a Deep Convolutional Neural Network (DCNN) based method. Firstly, we designed the OS detection model OSDTAU-Net based on the TransU-Net architecture, which achieves a balance between global perception and local focus by introducing the Transformer and Attention Gate mechanisms into the classical U-Net model. Secondly, we construct the Oil Spill PolSAR dataset from Sentinel-1 satellite images, which further enhances the detection capability of the model through multi-feature fusion of polarimetric information. The experimental results validate that the proposed OS detection model surpasses traditional supervised classification and classic semantic segmentation algorithms, achieving an overall accuracy of 97.09% and F1 score of 89.54%. Furthermore, using the proposed model, we successfully identified two oil spill incidents in the northern Gulf of Mexico. The generated oil spill distribution maps exhibit better continuity and closely align with the actual conditions.

Swin-Caption: Swin Transformer-Based Image Captioning with Feature Enhancement and Multi-Stage Fusion

The objective of image captioning is to empower computers to generate human-like sentences autonomously, describing a provided image. To tackle the challenges of insufficient accuracy in image feature extraction and underutilization of visual information, we present a Swin Transformer-based model for image captioning with feature enhancement and multi-stage fusion (Swin-Caption). Initially, the Swin Transformer is employed in the capacity of an encoder for extracting images, while feature enhancement is adopted to gather additional image feature information. Subsequently, a multi-stage image and semantic fusion module is constructed to utilize the semantic information from past time steps. Lastly, a two-layer LSTM is utilized to decode semantic and image data, generating captions. The proposed model outperforms the baseline model in experimental tests and instance analysis on the public datasets Flickr8K, Flickr30K, and MS-COCO.

Research on a fault-diagnosis strategy of lithium iron phosphate battery in an energy-storage system based on multi-feature fusion

Quickly and accurately detecting the voltage abnormality of lithium-ion batteries in battery energy storage systems (BESS) can avoid accidents caused by battery faults. A triple-layer battery fault diagnosis strategy is proposed based on multi-feature fusion. The threshold are mainly based on the dynamic quartile method and the adaptive multiple is automatically adjusted according to the data situation. The first layer fault is based on the frequency of voltage values exceeding the limit to detect over-charge or over-discharge faults. The features such as improved stack autoencoder, improved standard deviation and improved Shannon entropy of the voltage sequences in a sliding-time window are mutually validated to identify open circuit or short circuit fault batteries in the second layer. The above features are input into the sliding local outlier factor (SLOF) method in the third layer to obtain the fused outlier factors. The grades of inconsistency are divided by fused outlier factor and verify the fault diagnosis results of the first two layers. Finally, validation is carried out using battery data from two actual operating BESSs. Moreover, a comparative analysis with existing methods has been executed. This proposed method can determine the type and severity of battery faults, which is conducive to the practical fault diagnosis application of battery management system in BESSs.

Fault diagnosis for multi-axis carving machine systems with Gaussian mixture hidden Markov models: A data-model interactive perspective

This paper is concerned with sensor fault diagnosis problems for multi-axis carving machine systems (MACMSs) with repetitive machining tasks. A novel fault diagnosis method that combines the multi-feature fusion technology and Gaussian mixture hidden Markov models (GMHMMs) is proposed, which is inspired by a data- and model-driven collaborative perspective. With fault-sensitive features first extracted from both the time domain and time–frequency domain, the composite health index (CHI) is constructed to facilitate the understanding of the time-varying evolution. Then, GMHMMs are established to characterize the probabilistic relationship between the hidden states and CHI. To achieve high-precision fault classification, a well-designed global objective function is adopted to dynamically optimize both the CHI construction and classifier model training in a closed-loop feedback mechanism. Specifically, the fusion coefficients with range and equality constraints are integrated as part of the model parameters into the global optimization objective function, thereby reducing the search range and improving convergence speed. Besides, the well-trained GMHMMs interact with each other to capture the correlation information between different faults, and are utilized for online fault diagnosis. Finally, experiments are conducted on a self-developed multi-axis carving machine platform. The results exhibit outstanding performance in comparison with existing methods, particularly attaining a diagnostic accuracy of 95.37%.

A multi-feature fusion graph attention network for decoding motor imagery intention in spinal cord injury patients.

Electroencephalogram (EEG) signals exhibit multi-domain features, and electrode distributions follow non-Euclidean topology. To fully resolve the EEG signals, this study proposes a Temporal-Frequency-Spatial multi-domain feature fusion Graph Attention Network (TFSGAT) for motor imagery (MI) intention recognition in spinal cord injury (SCI) patients. The proposed model uses phase-locked value (PLV) to extract spatial phase connectivity information between EEG channels and continuous wavelet transform to extract valid EEG information in the time-frequency domain. It then models a graph data structure containing multi-domain information. The gated recurrent unit and GAT learn EEG's dynamic temporal-spatial information. Finally, the fully connected layer outputs the MI intention recognition results. After 10 times 10-fold cross-validation, the proposed model can achieve an average accuracy of 95.82%. Furthermore, this study analyzes the Event-Related Desynchronization/Event-Related Synchronization and PLV brain network to explore the brain activity of SCI patients during MI. This study confirms the potential of the proposed model in terms of EEG decoding performance and provides a reference for the mechanism of neural activity in SCI patients.

MuSE: A deep learning model based on multi-feature fusion for super-enhancer prediction

Although bioinformatics-based methods accurately identify SEs (Super-enhancers), the results depend on feature design. It is foundational to representing biological sequences and automatically extracting their key features for improving SE identification. We propose a deep learning model MuSE (Multi-Feature Fusion for Super-Enhancer), based on multi-feature fusion. This model utilizes two encoding methods, one-hot and DNA2Vec, to signify DNA sequences. Specifically, one-hot encoding reflects single nucleotide information, while k-mer representations based on DNA2Vec capture both local sequence fragment information and global sequence characteristics. These types of feature vectors are conducted and combined by neural networks, which aim at SE prediction. To validate the effectiveness of MuSE, we design extensive experiments on human and mouse species datasets. Compared to baselines such as SENet, MuSE improves the prediction of F1 score to a maximum improvement exceeding 0.05 on mouse species. The k-mer representations based on DNA2Vec among the given features have the most important impact on predictions. This feature effectively captures context semantic knowledge and positional information of DNA sequences. However, its representation of the individuality of each species negatively affects MuSE's generalization ability. Nevertheless, the cross-species prediction results of MuSE improve again to reach an AUC of nearly 0.8, after removing this type of feature. Source codes are available at https://github.com/15831959673/MuSE.

Rural Tourist Attractions Recommendation Model Based on Multi-Feature Fusion Graph Neural Networks

With the rapid growth of the rural tourism industry, traditional tourism recommendation technologies can no longer meet the necessary requirements. To address the issue of rural tourist attraction recommendations, a rural tourist attraction recommendation model is constructed based on a multi-feature fusion graph neural network. First, construct a feature map based on the relationship between tourists’ preferences and tourist attractions, and incorporate the attention mechanism to enhance the model’s learning capabilities. Second, utilize a two-part graph model to extract positive and negative preference features of tourists, and a conversation graph model to extract tourists’ transfer preference features. Finally, various features are utilized to generate suggested content by computing scores for tourists’ travel preferences. To address the problem of recommending tourist groups, suitable features for random group matching are collected and the cosine function is employed to identify users with similar random group features. Finally, the multi-features are merged, and the tourists’ interest preferences are scored to arrive at content recommendations. In the experiment on individualized attraction recommendations, data from the Chengdu area were used to test the proposed model. The accuracy of the model’s recommendations was 0.822 for five recommendations which outperformed the other models. In the experiment for group-based attraction recommendations, this experiment tested the Chengdu dataset. The proposed model achieved the highest accuracy of 0.972 when the group size was 70, outperforming the other two models. Additionally, with regards to different numbers of recommendations, the proposed model’s accuracy was 0.5241, which was the best performance among the three models when the number of recommendations was set to five. The proposed recommendation model performs optimally in suggesting tourist attractions and meets the needs of rural tourism. The research content provides crucial technical references for tourist traveling and rural tourism development.

Mf-net: multi-feature fusion network based on two-stream extraction and multi-scale enhancement for face forgery detection

Due to the increasing sophistication of face forgery techniques, the images generated are becoming more and more realistic and difficult for human eyes to distinguish. These face forgery techniques can cause problems such as fraud and social engineering attacks in facial recognition and identity verification areas. Therefore, researchers have worked on face forgery detection studies and have made significant progress. Current face forgery detection algorithms achieve high detection accuracy within-dataset. However, it is difficult to achieve satisfactory generalization performance in cross-dataset scenarios. In order to improve the cross-dataset detection performance of the model, this paper proposes a multi-feature fusion network based on two-stream extraction and multi-scale enhancement. First, we design a two-stream feature extraction module to obtain richer feature information. Secondly, the multi-scale feature enhancement module is proposed to focus the model more on information related to the current sub-region from different scales. Finally, the forgery detection module calculates the overlap between the features of the input image and real images during the training phase to determine the forgery regions. The method encourages the model to mine forgery features and learns generic and robust features not limited to a particular feature. Thus, the model achieves high detection accuracy and performance. We achieve the AUC of 99.70% and 90.71% on FaceForensics++ and WildDeepfake datasets. The generalization experiments on Celeb-DF-v2 and WildDeepfake datasets achieve the AUC of 80.16% and 65.15%. Comparison experiments with multiple methods on other benchmark datasets confirm the superior generalization performance of our proposed method while ensuring model detection accuracy. Our code can be found at: https://github.com/1241128239/MFNet.

Integrated Extraction of Entities and Relations via Attentive Graph Convolutional Networks

For information security, entity and relation extraction can be applied in sensitive information protection, data leakage detection, and other aspects. The current approaches to entity relation extraction not only ignore the relevance and dependency between name entity recognition and relation extraction but also may result in the cumulative propagation of errors. To solve this problem, it is proposed that an end-to-end joint entity and relation extraction model based on the Attention mechanism and Graph Convolutional Network (GCN) to simultaneously extract named entities and their relationships. The model includes three parts: the detection of entity span, the construction of an entity relation weighted graph, and the inference of entity relation type. Firstly, the detection of entity spans is viewed as a sequence labeling problem, and a multi-feature fusion approach for word embedding representation is designed to calculate all entity spans in a sentence to form an entity span matrix. Secondly, the entity span matrix is employed in the Multi-Head Attention mechanism for constructing the weighted adjacency matrix of the entity relation graph. Finally, for the inference of entity relation type, considering the interaction between entities and relations, the entity span matrix and relation connection matrix are simultaneously fed into the GCN for integrated extraction of entities and relations. Our model is evaluated on the public NYT dataset, attaining a precision of 66.4%, a recall of 63.1%, and an F1 score of 64.7% for joint entity and relation extraction, significantly outperforming other approaches. Experiments demonstrate that the proposed model is helpful for inferring entities and relations, considering the interaction between entities and relations through the Attention mechanism and GCN.

MSL-CCRN: Multi-stage self-supervised learning based cross-modality contrastive representation network for infrared and visible image fusion

Infrared and visible image fusion (IVIF) facing different information in two modal scenarios, the focus of research is to better extract different information. In this work, we propose a multi-stage self-supervised learning based cross-modality contrastive representation network for infrared and visible image fusion (MSL-CCRN). Firstly, considering that the scene differences between different modalities affect the fusion of cross-modal images, we propose a contrastive representation network (CRN). CRN enhances the interaction between the fused image and the source image, and significantly improves the similarity between the meaningful features in each modality and the fused image. Secondly, due to the lack of ground truth in IVIF, the quality of directly obtained fused image is seriously affected. We design a multi-stage fusion strategy to address the loss of important information in this process. Notably, our method is a self-supervised network. In fusion stage I, we reconstruct the initial fused image as the new view of fusion stage II. In fusion stage II, we use the fused image obtained in the previous stage to carry out three-view contrastive representation, thereby constraining the feature extraction of the source image. This makes the final fused image introduce more important information in the source image. Through a large number of qualitative, quantitative experiments and downstream object detection experiments, our propose method shows excellent performance compared with most advanced methods.

Cattle identification based on multiple feature decision layer fusion

In farming scenarios, cattle identification has become a key issue for the development of precision farming. In precision livestock farming, single-feature recognition methods are prone to misjudgment in complex scenarios involving multiple cattle obscuring each other during drinking and feeding. This paper proposes a decision-level identification method based on the multi-feature fusion of cattle faces, muzzle patterns, and ear tags. The method utilizes the SOLO algorithm to segment images and employs the FaceNet and PP-OCRv4 networks to extract features for the cattle’s faces, muzzle patterns, and ear tags. These features are compared with the Ground truth, from which the Top 3 features are extracted. The corresponding cattle IDs of these features are then processed using One-Hot encoding to serve as the final input for the decision layer, and various ensemble strategies are used to optimize the model. The results show that using the multimodal decision fusion method makes the recognition accuracy reach 95.74%, 1.4% higher than the traditional optimal unimodal recognition accuracy. The verification rate reaches 94.72%, 10.65% higher than the traditional optimal unimodal recognition verification rate. The research results demonstrate that the multi-feature fusion recognition method has significant advantages in drinking and feeding farm environments, providing an efficient and reliable solution for precise identification and management of cattle in farms and significantly improving recognition accuracy and stability.

Local-Global Feature Adaptive Fusion Network for Building Crack Detection.

Cracks represent one of the most common types of damage in building structures and it is crucial to detect cracks in a timely manner to maintain the safety of the buildings. In general, tiny cracks require focusing on local detail information while complex long cracks and cracks similar to the background require more global features for detection. Therefore, it is necessary for crack detection to effectively integrate local and global information. Focusing on this, a local-global feature adaptive fusion network (LGFAF-Net) is proposed. Specifically, we introduce the VMamba encoder as the global feature extraction branch to capture global long-range dependencies. To enhance the ability of the network to acquire detailed information, the residual network is added as another local feature extraction branch, forming a dual-encoding network to enhance the performance of crack detection. In addition, a multi-feature adaptive fusion (MFAF) module is proposed to integrate local and global features from different branches and facilitate representative feature learning. Furthermore, we propose a building exterior wall crack dataset (BEWC) captured by unmanned aerial vehicles (UAVs) to evaluate the performance of the proposed method used to identify wall cracks. Other widely used public crack datasets are also utilized to verify the generalization of the method. Extensive experiments performed on three crack datasets demonstrate the effectiveness and superiority of the proposed method.

Detecting Water Stress in Winter Wheat Based on Multifeature Fusion from UAV Remote Sensing and Stacking Ensemble Learning Method

The integration of remote sensing technology and machine learning algorithms represents a new research direction for the rapid and large-scale detection of water stress in modern agricultural crops. However, in solving practical agricultural problems, single machine learning algorithms cannot fully explore the potential information within the data, lacking stability and accuracy. Stacking ensemble learning (SEL) can combine the advantages of multiple single machine learning algorithms to construct more stable predictive models. In this study, threshold values of stomatal conductance (gs) under different soil water stress indices (SWSIs) were proposed to assist managers in irrigation scheduling. In the present study, six irrigation treatments were established for winter wheat to simulate various soil moisture supply conditions. During the critical growth stages, gs was measured and the SWSI was calculated. A spectral camera mounted on an unmanned aerial vehicle (UAV) captured reflectance images in five bands, from which vegetation indices and texture information were extracted. The results indicated that gs at different growth stages of winter wheat was sensitive to soil moisture supply conditions. The correlation between the gs value and SWSI value was high (R2 > 0.79). Therefore, the gs value threshold can reflect the current soil water stress level. Compared with individual machine learning models, the SEL model exhibited higher prediction accuracy, with R2 increasing by 6.67–17.14%. Using a reserved test set, the SEL model demonstrated excellent performance in various evaluation metrics across different growth stages (R2: 0.69–0.87, RMSE: 0.04–0.08 mol m−2 s−1; NRMSE: 12.3–23.6%, MAE: 0.03–0.06 mol m−2 s−1) and exhibited excellent stability and accuracy. This research can play a significant role in achieving large-scale monitoring of crop growth status through UAV, enabling the real-time capture of winter wheat water deficit changes, and providing technical support for precision irrigation.

Low-Complexity Wireless Technique Classification With Multifeature Fusion Broad Learning Network

Low-Complexity Wireless Technique Classification With Multifeature Fusion Broad Learning Network

State of Health Estimation Method for Lithium-Ion Batteries Based on Multifeature Fusion and BO-BiGRU Model

Abstract The state of health (SOH) of lithium-ion batteries is a crucial parameter for assessing battery degradation. The aim of this study is to solve the problems of single extraction of health features (HFs) and redundancy of information between features in the SOH estimation. This article develops an SOH estimation method for lithium-ion batteries based on multifeature fusion and Bayesian optimization (BO)-bidirectional gated recurrent unit (BiGRU) model. First, a total of eight HFs in three categories, namely, time, energy, and probability, can be extracted from the charging data to accurately describe the aging mechanism of the battery. The Pearson and Spearman analysis method verified the strong correlation between HFs and SOH. Second, the multiple principal components obtained by kernel principal component analysis (KPCA) can eliminate the redundancy of information between HFs. The principal component with the highest correlation with SOH is selected by bicorrelation analysis to be defined as the fused HF. Finally, to improve SOH estimation accuracy, the BO-BiGRU model is proposed. The proposed method is validated using battery datasets from NASA. The results show that the SOH estimation accuracy of the BO-BiGRU model proposed in this article is high, while mean absolute error (MAE) is lower than 1.2%. In addition, the SOH of the lithium battery is estimated using different proportions of test sets, and the results show that the root-mean-square error (RMSE) and the mean absolute percentage error (MAPE) of the SOH remain within 3%, with high estimation accuracy and robustness.

Research on registration method for enface image using multi-feature fusion

Objective.The purpose of this work is to accurately and quickly register the Optical coherence tomography (OCT) projection (enface) images at adjacent time points, and to solve the problem of interference caused by CNV lesions on the registration features.Approach.In this work, a multi-feature registration strategy was proposed, in which a combined feature (com-feature) containing 3D information, intersection information and SURF feature was designed. Firstly, the coordinates of all feature points were extracted as combined features, and then these feature coordinates were added to the initial vascular coordinate set simplified by the Douglas-Peucker algorithm as the point set for registration. Finally, the coherent point drift registration algorithm was used to register the enface coordinate point sets of adjacent time series.Main results.The newly designed features significantly improve the success rate of global registration of vascular networks in enface images, while the simplification step greatly improves the registration speed on the basis of preserving vascular features. The MSE, DSC and time complexity of the proposed method are 0.07993, 0.9693 and 42.7016 s, respectively.Significance.CNV is a serious retinal disease in ophthalmology. The registration of OCT enface images at adjacent time points can timely monitor the progress of the disease and assist doctors in making diagnoses. The proposed method not only improves the accuracy of OCT enface image registration, but also significantly reduces the time complexity. It has good registration results in clinical routine and provides a more efficient method for clinical diagnosis and treatment.

SPFDNet: Water Extraction Method Based on Spatial Partition and Feature Decoupling

Extracting water information from remote-sensing images is of great research significance for applications such as water resource protection and flood monitoring. Current water extraction methods aggregated richer multi-level features to enhance the output results. In fact, there is a difference in the requirements for the water body and the water boundary. Indiscriminate multi-feature fusion can lead to perturbation and competition of information between these two types of features during the optimization. Consequently, models cannot accurately locate the internal vacancies within the water body with the external boundary. Therefore, this paper proposes a water feature extraction network with spatial partitioning and feature decoupling. To ensure that the water features are extracted with deep semantic features and stable spatial information before decoupling, we first design a chunked multi-scale feature aggregation module (CMFAM) to construct a context path for obtaining deep semantic information. Then, an information interaction module (IIM) is designed to exchange information between two spatial paths with two fixed resolution intervals and the two paths through. During decoding, a feature decoupling module (FDM) is developed to utilize internal flow prediction to acquire the main body features, and erasing techniques are employed to obtain boundary features. Therefore, the deep features of the water body and the detailed boundary information are supplemented, strengthening the decoupled body and boundary features. Furthermore, the integrated expansion recoupling module (IERM) module is designed for the recoupling stage. The IERM expands the water body and boundary features using expansion and adaptively compensates the transition region between the water body and boundary through information guidance. Finally, multi-level constraints are combined to realize the supervision of the decoupled features. Thus, the water body and boundaries can be extracted more accurately. A comparative validation analysis is conducted on the public datasets, including the gaofen image dataset (GID) and the gaofen2020 challenge dataset (GF2020). By comparing with seven SOTAs, the results show that the proposed method achieves the best results, with IOUs of 91.22 and 78.93, especially in the localization of water bodies and boundaries. By applying the proposed method in different scenarios, the results show the stable capability of the proposed method for extracting water with various shapes and areas.

Adaptive feature fusion and disturbance correction for accurate remaining useful life prediction of rolling bearings

As a key component in the transmission system of high-speed trains, bearings need to withstand certain loads while rotating at high speeds. Once a failure occurs, it can directly affect the safety of train operations. Therefore, it is of great significance to establish a reliable remaining useful life model to ensure train operation safety. Addressing the problems of redundant features in the existing multi-feature fusion process, which affects diagnostic performance, and the spurious fluctuations in the fusion features, which cause inaccurate determination of the start fault time and lead to low prediction accuracy of remaining useful life, we propose a method based on adaptive feature fusion and the autoregressive integrated moving average model for rolling bearing prediction. Firstly, features from the time domain, frequency domain, and entropy are extracted. A feature selection mechanism with minimum redundancy is constructed to screen the optimal sensitive feature set. Secondly, based on adaptive feature fusion, the optimal sensitive feature set is dynamically fused, and the spurious fluctuations of the health index are corrected using linear regression and the 3σ principle. Next, a bottom-up time series segmentation method is employed to divide the health status of the Improved Health Indicators. Finally, a remaining useful life prediction model based on the autoregressive integrated moving average model is established. This study demonstrates that the proposed method effectively identifies features that are most sensitive to degradation trends, accurately determines the initial fault moment of bearings, and achieves effective prediction of the remaining useful life of bearings.