Final Feature Research Articles

SFT-Net: A Network for Detecting Fatigue From EEG Signals by Combining 4D Feature Flow and Attention Mechanism.

Fatigued driving is a leading cause of traffic accidents, and accurately predicting driver fatigue can significantly reduce their occurrence. However, modern fatigue detection models based on neural networks often face challenges such as poor interpretability and insufficient input feature dimensions. This article proposes a novel Spatial-Frequency-Temporal Network (SFT-Net) method for detecting driver fatigue using electroencephalogram (EEG) data. Our approach integrates EEG signals' spatial, frequency, and temporal information to improve recognition performance. We transform the differential entropy of five frequency bands of EEG signals into a 4D feature tensor to preserve these three types of information. An attention module is then used to recalibrate the spatial and frequency information of each input 4D feature tensor time slice. The output of this module is fed into a depthwise separable convolution (DSC) module, which extracts spatial and frequency features after attention fusion. Finally, long short-term memory (LSTM) is used to extract the temporal dependence of the sequence, and the final features are output through a linear layer. We validate the effectiveness of our model on the SEED-VIG dataset, and experimental results demonstrate that SFT-Net outperforms other popular models for EEG fatigue detection. Interpretability analysis supports the claim that our model has a certain level of interpretability. Our work addresses the challenge of detecting driver fatigue from EEG data and highlights the importance of integrating spatial, frequency, and temporal information.

DSFE: Decoding EEG-Based Finger Motor Imagery Using Feature-Dependent Frequency, Feature Fusion and Ensemble Learning.

Accurate decoding finger motor imagery is essential for fine motor control using EEG signals. However, decoding finger motor imagery is particularly challenging compared with ordinary motor imagery. This paper proposed a novel EEG decoding method of feature-dependent frequency band selection, feature fusion, and ensemble learning (DSFE) for finger motor imagery. First, a feature-dependent frequency band selection method based on correlation coefficient (FDCC) was proposed to select feature-specific effective bands. Second, a feature fusion method was proposed to fuse different types of candidate features to produce multiple refined sets of decoding features. Finally, an ensemble model using the weighted voting strategy was proposed to make full use of these diverse sets of final features. The results on a public EEG dataset of five fingers motor imagery showed that the DSFE method is effective and achieves the highest decoding accuracy of 50.64%, which is 7.64% higher than existing studies using exactly the same data. The experiments further revealed that both the effective frequency bands of different subjects and the effective frequency bands of different types of features are different in finger motor imagery. Furthermore, compared with two-hand motor imagery, the effective decoding information of finger motor imagery is transferred to the lower frequency. The idea and findings in this paper provide a valuable perspective for understanding fine motor imagery in-depth.

Exploring Feature Selection and Classification Techniques to Improve the Performance of an Electroencephalography-Based Motor Imagery Brain-Computer Interface System.

The accuracy of classifying motor imagery (MI) activities is a significant challenge when using brain-computer interfaces (BCIs). BCIs allow people with motor impairments to control external devices directly with their brains using electroencephalogram (EEG) patterns that translate brain activity into control signals. Many researchers have been working to develop MI-based BCI recognition systems using various time-frequency feature extraction and classification approaches. However, the existing systems still face challenges in achieving satisfactory performance due to large amount of non-discriminative and ineffective features. To get around these problems, we suggested a multiband decomposition-based feature extraction and classification method that works well, along with a strong feature selection method for MI tasks. Our method starts by splitting the preprocessed EEG signal into four sub-bands. In each sub-band, we then used a common spatial pattern (CSP) technique to pull out narrowband-oriented useful features, which gives us a high-dimensional feature vector. Subsequently, we utilized an effective feature selection method, Relief-F, which reduces the dimensionality of the final features. Finally, incorporating advanced classification techniques, we classified the final reduced feature vector. To evaluate the proposed model, we used the three different EEG-based MI benchmark datasets, and our proposed model achieved better performance accuracy than existing systems. Our model's strong points include its ability to effectively reduce feature dimensionality and improve classification accuracy through advanced feature extraction and selection methods.

Feature boosting with efficient attention for scene parsing

The complexity of scene parsing grows with the number of object and scene classes, which is higher in unrestricted open scenes. The biggest challenge is to model the spatial relation between scene elements while succeeding in identifying objects at smaller scales. This paper presents a novel feature-boosting network that gathers spatial context from multiple levels of feature extraction and computes the attention weights for each level of representation to generate the final class labels. A novel ‘channel attention module’ is designed to compute the attention weights, ensuring that features from the relevant extraction stages are boosted while the others are attenuated. The model also learns spatial context information at low resolution to preserve the abstract spatial relationships among scene elements and reduce computational cost. Spatial attention is subsequently concatenated into a final feature set before applying feature boosting. Low-resolution spatial attention features are trained using an auxiliary task that help to learn a coarse global scene structure. The proposed model outperforms all state-of-the-art models on both the ADE20K and the Cityscapes datasets.

Semi-Identical Twins Variational AutoEncoder for Few-Shot Learning.

Data augmentation is a popular way for few-shot learning (FSL). It generates more samples as supplements and then transforms the FSL task into a common supervised learning problem for a solution. However, most data-augmentation-based FSL approaches only consider the prior visual knowledge for feature generation, thereby leading to low diversity and poor quality of generated data. In this study, we attempt to address this issue by incorporating both prior visual and prior semantic knowledge to condition the feature generation process. Inspired by some genetic characteristics of semi-identical twins, a novel multimodal generative FSL approach was developed named semi-identical twins variational autoencoder (STVAE) to better exploit the complementarity of these modality information by considering the multimodal conditional feature generation process as a process that semi-identical twins are born and collaborate to simulate their father. STVAE conducts feature synthesis by pairing two conditional variational autoencoders (CVAEs) with the same seed but different modality conditions. Subsequently, the generated features of two CVAEs are considered as semi-identical twins and adaptively combined to yield the final feature, which is considered as their fake father. STVAE requires that the final feature can be converted back into its paired conditions while ensuring these conditions remain consistent with the original in both representation and function. Moreover, STVAE is able to work in the partial modality-absence case due to the adaptive linear feature combination strategy. STVAE essentially provides a novel idea to exploit the complementarity of different modality prior information inspired by genetics in FSL. Extensive experimental results demonstrate that our work achieves promising performances in comparison to the recent state-of-the-art approaches, as well as validate its effectiveness on FSL under various modality settings.

Insights on the role of blocking agent on the properties of the lipase from Thermomyces lanuginosus immobilized on heterofunctional support for hydroesterification reactions

Here, we report a study of the effect of the blocking agent on the properties of the lipase from Thermomyces lanuginosus (TLL) immobilized on a heterofunctional support (Purolite C18-ethylnediamina (EDA)- vinyl sulfone (VS)-TLL-blocking agent) in different reactions. The performance of the biocatalysts was compared to those immobilized on standard hydrophobic support (Purolite C18-TLL) and the commercial one (TLL-IM). The nature of the blocking agent (Cys, Gly and Asp) altered the enzyme features. TLL-IM always gave a comparatively worse performance, with its specificity for the oil being very different to the Purolite biocatalysts. Under optimized conditions, Purolite C18-TLL yielded 97 % of hydrolysis conversion after 4 h using a water/waste cooking soybean oil (WCSO) mass ratio of 4.3, biocatalyst load of 6.5 wt% and a temperature of 44.2 °C (without buffer or emulsification agent). In esterification reactions of the purified free fatty acids (FFAs) obtained from WCSO, the best TLL biocatalysts depended on the utilized alcohol: linear amyl alcohol was preferred by Purolite C18-TLL and Purolite C18-EDA-VS-TLL-Gly, while higher activity was achieved utilizing isoamyl alcohol as nucleophile by Purolite C18-EDA-VS-TLL-Cys, Purolite C18-EDA-VS-TLL-Asp and IM-TLL as catalysts. All the results indicate the influence of the blocking step on the final biocatalyst features.

Fragment-Fusion Transformer: Deep Learning-Based Discretization Method for Continuous Single-Cell Raman Spectral Analysis.

Raman spectroscopy has become an important single-cell analysis tool for monitoring biochemical changes at the cellular level. However, Raman spectral data, typically presented as continuous data with high-dimensional characteristics, is distinct from discrete sequences, which limits the application of deep learning-based algorithms in data analysis due to the lack of discretization. Herein, a model called fragment-fusion transformer is proposed, which integrates the discrete fragmentation of continuous spectra based on their intrinsic characteristics with the extraction of intrafragment features and the fusion of interfragment features. The model integrates the intrinsic feature-based fragmentation of spectra with transformer, constructing the fragment transformer block for feature extraction within fragments. Interfragment information is combined through the pyramid design structure to improve the model's receptive field and fully exploit the spectral properties. During the pyramidal fusion process, the information gain of the final extracted features in the spectrum has been enhanced by a factor of 9.24 compared to the feature extraction stage within the fragment, and the information entropy has been enhanced by a factor of 13. The fragment-fusion transformer achieved a spectral recognition accuracy of 94.5%, which is 4% higher compared to the method without fragmentation and fusion processes on the test set of cell Raman spectroscopy identification experiments. In comparison to common spectral classification models such as KNN, SVM, logistic regression, and CNN, fragment-fusion transformer has achieved 4.4% higher accuracy than the best-performing CNN model. Fragment-fusion transformer method has the potential to serve as a general framework for discretization in the field of continuous spectral data analysis and as a research tool for analyzing the intrinsic information within spectra.

Interactions among dimensions of divergent thinking as predictors of creative activity and accomplishment

Divergent thinking (DT) tests are often used to estimate creative potential. They have sound theoretical bases, good reliability, and moderate predictive validity. One previous study that sampled gifted and nongifted children tested the interactions among DT indices (fluency, originality, and flexibility) in the predictive equation and reported a respectable validity coefficient of 0.59. Two meta-analyses have also supported the use of DT tests as estimates of creative potential, but neither of these examined interactions among the DT indices. The present correlational, group-comparison (N = 312) investigation (a) replicated the 1986 study and examined interactions among the three DT indices and (b) extended that research by using two newer criterion measures. One was the Creative Activity and Accomplishment Checklist (CAAC) which had been updated such that it included scales for Technological, Athletic, and Everyday creativity, in addition to scales for Art, Science & Math, Writing, and Music. The updated CAAC also used a new method for assessing the quality of creative accomplishment (in addition to the quantity of creative activity). A final novel feature of the present investigation was the use of a second criterion, one that was developed specifically for research on DT. Hierarchical regression analyses indicated that the fluency X originality and flexibility X originality interactions were significantly related to the Quantity of creativity activity score. Giftedness, fluency, and the fluency X originality interaction were significantly related to the Quality score. Canonical correlation analyses indicated a predictive validity of 0.73 (p < 0.001) for the total sample. One clear conclusion is therefore that predictions from DT tests should take interactions into account. These findings are discussed in the context of theories which recognize creativity as an important part of giftedness and theories which suggest that several facets of DT need to be taken into account when assessing creative cognition.

MFSC: A Multimodal Aspect-Level Sentiment Classification Framework with Multi-Image Gate and Fusion Networks

Currently, there is a great deal of interest in multimodal aspect-level sentiment classification using both textual and visual information, which changes the traditional use of only single-modal to identify sentiment polarity. Considering that existing methods could be strengthened in terms of classification accuracy, we conducted a study on aspect-level multimodal sentiment classification with the aim of exploring the interaction between textual and visual features. Specifically, we construct a multimodal aspect-level sentiment classification framework with multi-image gate and fusion networks called MFSC. MFSC consists of four parts, i.e., text feature extraction, visual feature extraction, text feature enhancement, and multi-feature fusion. Firstly, a bidirectional long short-term memory network is adopted to extract the initial text feature. Based on this, a text feature enhancement strategy is designed, which uses text memory network and adaptive weights to extract the final text features. Meanwhile, a multi-image gate method is proposed for fusing features from multiple images and filtering out irrelevant noise. Finally, a text-visual feature fusion method based on an attention mechanism is proposed to better improve the classification performance by capturing the association between text and images. Experimental results show that MFSC has advantages in classification accuracy and macro-F1.

Fusing differentiable rendering and language–image contrastive learning for superior zero-shot point cloud classification

Zero-shot point cloud classification involves recognizing categories not encountered during training. Current models often exhibit reduced accuracy on unseen categories without 3D pre-training, emphasizing the need for improved precision and interoperability. We propose a novel approach integrating differentiable rendering with contrastive language–image pre-training. Initially, differentiable rendering autonomously learns representative viewpoints from the data, enabling the transformation of point clouds into multi-view images while preserving key visual information. This transformation facilitates optimized viewpoint selection during training, refining the final feature representation. Features are extracted from the multi-view images and integrated into a global multi-view feature using a cross-attention mechanism. On the textual side, a large language model (LLM) is provided with 3D heuristic prompts to generate 3D-specific text reflecting category-specific traits, from which textual features are derived. The LLM’s extensive pre-trained knowledge enables it to capture abstract notions and categorical features relevant to distinct point cloud categories. Visual and textual features are aligned in a unified embedding space, enabling zero-shot classification. Throughout training, the Structural Similarity Index (SSIM) is integrated into the loss function to encourage the model to discern more distinctive viewpoints, reduce redundancy in multi-view imagery, and enhance computational efficiency. Experimental results on the ModelNet10, ModelNet40, and ScanObjectNN datasets demonstrate classification accuracies of 75.68%, 66.42%, and 52.03%, respectively, surpassing prevailing methods in zero-shot point cloud classification accuracy.

Test Input Prioritization for 3D Point Clouds

3D point cloud applications have become increasingly prevalent in diverse domains, showcasing their efficacy in various software systems. However, testing such applications presents unique challenges due to the high-dimensional nature of 3D point cloud data and the vast number of possible test cases. Test input prioritization has emerged as a promising approach to enhance testing efficiency by prioritizing potentially misclassified test cases during the early stages of the testing process. Consequently, this enables the early labeling of critical inputs, leading to a reduction in the overall labeling cost. However, applying existing prioritization methods to 3D point cloud data is constrained by several factors: (1) inadequate consideration of crucial spatial information, and (2) susceptibility to noises inherent in 3D point cloud data. In this article, we propose PCPrior , the first test prioritization approach specifically designed for 3D point cloud test cases. The fundamental concept behind PCPrior is that test inputs closer to the decision boundary of the model are more likely to be predicted incorrectly. To capture the spatial relationship between a point cloud test and the decision boundary, we propose transforming each test (a point cloud) into a low-dimensional feature vector, toward indirectly revealing the underlying proximity between a test and the decision boundary. To achieve this, we carefully design a group of feature generation strategies, and for each test input, we generate four distinct types of features, namely spatial features, mutation features, prediction features, and uncertainty features. Through a concatenation of the four feature types, PCPrior assembles a final feature vector for each test. Subsequently, a ranking model is employed to estimate the probability of misclassification for each test based on its feature vector. Finally, PCPrior ranks all tests based on their misclassification probabilities. We conducted an extensive study based on 165 subjects to evaluate the performance of PCPrior, encompassing both natural and noisy datasets. The results demonstrate that PCPrior outperforms all of the compared test prioritization approaches, with an average improvement of 10.99% to 66.94% on natural datasets and 16.62% to 53% on noisy datasets.

Combining machine and deep transfer learning for mediastinal lymph node evaluation in patients with lung cancer

BackgroundThe prognosis and survival of patients with lung cancer are likely to deteriorate with metastasis. Using deep-learning in the detection of lymph node metastasis can facilitate the noninvasive calculation of the likelihood of such metastasis, thereby providing clinicians with crucial information to enhance diagnostic precision and ultimately improve patient survival and prognosis MethodsIn total, 623 eligible patients were recruited from two medical institutions. Seven deep learning models, namely Alex, GoogLeNet, Resnet18, Resnet101, Vgg16, Vgg19, and MobileNetv3 (small), were utilized to extract deep image histological features. The dimensionality of the extracted features was then reduced using the Spearman correlation coefficient (r ≥ 0.9) and Least Absolute Shrinkage and Selection Operator. Eleven machine learning methods, namely Support Vector Machine, K-nearest neighbor, Random Forest, Extra Trees, XGBoost, LightGBM, Naive Bayes, AdaBoost, Gradient Boosting Decision Tree, Linear Regression, and Multilayer Perceptron, were employed to construct classification prediction models for the filtered final features. The diagnostic performances of the models were assessed using various metrics, including accuracy, area under the receiver operating characteristic curve, sensitivity, specificity, positive predictive value, and negative predictive value. Calibration and decision-curve analyses were also performed. ResultsThe present study demonstrated that using deep radiomic features extracted from Vgg16, in conjunction with a prediction model constructed via a linear regression algorithm, effectively distinguished the status of mediastinal lymph nodes in patients with lung cancer. The performance of the model was evaluated based on various metrics, including accuracy, area under the receiver operating characteristic curve, sensitivity, specificity, positive predictive value, and negative predictive value, which yielded values of 0.808, 0.834, 0.851, 0.745, 0.829, and 0.776, respectively. The validation set of the model was assessed using clinical decision curves, calibration curves, and confusion matrices, which collectively demonstrated the model's stability and accuracy ConclusionIn this study, information on the deep radiomics of Vgg16 was obtained from computed tomography images, and the linear regression method was able to accurately diagnose mediastinal lymph node metastases in patients with lung cancer.

Investigating the sensitivity of dual-phase steel microstructure to continuous annealing line parameters

The mechanical properties of dual-phase (DP) steels are influenced by ferrite grain size, martensite distribution/morphology, and strength differences between ferrite/martensite phases. A sensitivity study was conducted on a continuous annealing line cycle of DP steel manufacturing process to understand the effect of individual segments on the final microstructural features. From the study results, an optimised cycle was developed to obtain a DP steel microstructure with reduced martensite banding, fine ferrite grain size and less untempered martensite. The main modifications are austenite soaking temperature above Ac3 temperature, higher cooling rate of the slow cooling stage, increased soak time at ferrite soaking temperature, and higher over-aging temperature. The optimised cycle produced DP steel microstructure with reduced martensite banding at a similar phase fraction (35%) and decreased ferrite grain size from 8.4 to 5.1 µm. This improved the mechanical properties, increasing tensile strength from 789 ± 21 to 831 ± 14 MPa and total elongation from 15 ± 3 to 20 ± 2%.

Biomedical named entity recognition based on multi-cross attention feature fusion.

Currently, in the field of biomedical named entity recognition, CharCNN (Character-level Convolutional Neural Networks) or CharRNN (Character-level Recurrent Neural Network) is typically used independently to extract character features. However, this approach does not consider the complementary capabilities between them and only concatenates word features, ignoring the feature information during the process of word integration. Based on this, this paper proposes a method of multi-cross attention feature fusion. First, DistilBioBERT and CharCNN and CharLSTM are used to perform cross-attention word-char (word features and character features) fusion separately. Then, the two feature vectors obtained from cross-attention fusion are fused again through cross-attention to obtain the final feature vector. Subsequently, a BiLSTM is introduced with a multi-head attention mechanism to enhance the model's ability to focus on key information features and further improve model performance. Finally, the output layer is used to output the final result. Experimental results show that the proposed model achieves the best F1 values of 90.76%, 89.79%, 94.98%, 80.27% and 88.84% on NCBI-Disease, BC5CDR-Disease, BC5CDR-Chem, JNLPBA and BC2GM biomedical datasets respectively. This indicates that our model can capture richer semantic features and improve the ability to recognize entities.

Predictive Modeling for Functional Outcomes After All-Inside Arthroscopic Anterior Talofibular Ligament Repair of Chronic Ankle Instability.

Arthroscopic anterior talofibular ligament repair (AATFLR) is a surgical strategy to treat chronic ankle instability (CAI) patients. This study identified risk factors that influenced the functional outcomes of AATFLR for CAI and developed prognostic nomogram for predicting functional outcomes in future AATFLR cases. Patients undergoing AATFLR from January 2016 to June 2022 with at least 10 months of follow-up were included in the study. The Karlsson Ankle Functional Score (KAFS) was evaluated preoperatively and at last follow-up visit. A total of 15 potential predictors including age, sex, body mass index, side affected, time from injury to surgery, sports-related injury, osteophyte, loose bodies, distal tibiofibular syndesmosis, ATFL avulsion fracture, Outerbridge classification of osteochondral lesions, postoperative immobilization method, ambulation time, walking time, and follow-up time, were recorded. We first used univariate binary logistic regression analysis to select the potential significant prognostic features, which were then subjected to the least absolute shrinkage and selection operator (LASSO) regression algorithm for final feature selection. A nomogram based on the regression model was developed to estimate the functional outcomes of patients. Models were validated internally using bootstrapping and externally by calculating their performance on a validation cohort. Overall, 200 ankles fit inclusion criteria. Of these 200, a total of 185 (92.5%) ankles were eligible and divided into development (n = 121) and validation (n = 64) cohorts. Four predictors were ultimately included in the prognostic nomogram model: age, sex, sports-related injury, and postoperative immobilization method. We found in our cohort that the significant predictors of poorer functional outcomes of AATFLR were postoperative immobilization with lower-leg cast, female sex, non-sports-related ankle sprain, and increasing age. Prognostic nomograms were created.

PMSNet: Multiscale Partial-Discharge Signal Feature Recognition Model via a Spatial Interaction Attention Mechanism.

Partial discharge (PD) is a localized discharge phenomenon in the insulator of electrical equipment resulting from the electric field strength exceeding the local dielectric breakdown electric field. Partial-discharge signal identification is an important means of assessing the insulation status of electrical equipment and critical to the safe operation of electrical equipment. The identification effect of traditional methods is not ideal because the PD signal collected is subject to strong noise interference. To overcome noise interference, quickly and accurately identify PD signals, and eliminate potential safety hazards, this study proposes a PD signal identification method based on multiscale feature fusion. The method improves identification efficiency through the multiscale feature fusion and feature aggregation of phase-resolved partial-discharge (PRPD) diagrams by using PMSNet. The whole network consists of three parts: a CNN backbone composed of a multiscale feature fusion pyramid, a down-sampling feature enhancement (DSFB) module for each layer of the pyramid to acquire features from different layers, a Transformer encoder module dominated by a spatial interaction-attention mechanism to enhance subspace feature interactions, a final categorized feature recognition method for the PRPD maps and a final classification feature generation module (F-Collect). PMSNet improves recognition accuracy by 10% compared with traditional high-frequency current detection methods and current pulse detection methods. On the PRPD dataset, the validation accuracy of PMSNet is above 80%, the validation loss is about 0.3%, and the training accuracy exceeds 85%. Experimental results show that the use of PMSNet can greatly improve the recognition accuracy and robustness of PD signals and has good practicality and application prospects.

The Ins and Outs of Homeodomain-Leucine Zipper/Hormone Networks in the Regulation of Plant Development.

The generation of complex plant architectures depends on the interactions among different molecular regulatory networks that control the growth of cells within tissues, ultimately shaping the final morphological features of each structure. The regulatory networks underlying tissue growth and overall plant shapes are composed of intricate webs of transcriptional regulators which synergize or compete to regulate the expression of downstream targets. Transcriptional regulation is intimately linked to phytohormone networks as transcription factors (TFs) might act as effectors or regulators of hormone signaling pathways, further enhancing the capacity and flexibility of molecular networks in shaping plant architectures. Here, we focus on homeodomain-leucine zipper (HD-ZIP) proteins, a class of plant-specific transcriptional regulators, and review their molecular connections with hormonal networks in different developmental contexts. We discuss how HD-ZIP proteins emerge as key regulators of hormone action in plants and further highlight the fundamental role that HD-ZIP/hormone networks play in the control of the body plan and plant growth.

MASMDDI: multi-layer adaptive soft-mask graph neural network for drug-drug interaction prediction.

Accurately predicting Drug-Drug Interaction (DDI) is a critical and challenging aspect of the drug discovery process, particularly in preventing adverse reactions in patients undergoing combination therapy. However, current DDI prediction methods often overlook the interaction information between chemical substructures of drugs, focusing solely on the interaction information between drugs and failing to capture sufficient chemical substructure details. To address this limitation, we introduce a novel DDI prediction method: Multi-layer Adaptive Soft Mask Graph Neural Network (MASMDDI). Specifically, we first design a multi-layer adaptive soft mask graph neural network to extract substructures from molecular graphs. Second, we employ an attention mechanism to mine substructure feature information and update latent features. In this process, to optimize the final feature representation, we decompose drug-drug interactions into pairwise interaction correlations between the core substructures of each drug. Third, we use these features to predict the interaction probabilities of DDI tuples and evaluate the model using real-world datasets. Experimental results demonstrate that the proposed model outperforms state-of-the-art methods in DDI prediction. Furthermore, MASMDDI exhibits excellent performance in predicting DDIs of unknown drugs in two tasks that are more aligned with real-world scenarios. In particular, in the transductive scenario using the DrugBank dataset, the ACC and AUROC and AUPRC scores of MASMDDI are 0.9596, 0.9903, and 0.9894, which are 2% higher than the best performing baseline.

A texture feature extraction method considering spatial continuity and gray diversity

Texture features play an important role in the field of remote sensing classification. However, most existing methods lack a comprehensive consideration of spatial continuity, which makes them either destroy the spatial integrity of regular ground objects or fail to quantify the fragmentation degrees of irregular ground objects. These problems weak the ability of existing methods to distinguish ground objects with different fragmentation degrees. Therefore, this study proposed a new texture feature extraction method considering spatial continuity and gray diversity (SCGD). SCGD first connected all pixels in a neighborhood in series from end to end according to the row and column directions, and the diversities of the spatial continuity encoding in different directions were calculated by the Shannon index. Then, the Shannon index was used to calculate the gray diversity. Finally, SCGD calculated the weighted average of spatial continuity diversity and gray diversity to obtain the final texture feature values. Validation results indicated that SCGD can effectively distinguish ground objects with different fragmentation degrees, and its performance is better than that of traditional methods. As the spatial resolution decreases, its performance advantage becomes more obvious. Moreover, SCGD has great application potential in the field of ground object classification, and combining it with deep learning models will contribute to achieving the fine recognition of ground objects.

Double reuses based residual network

Deep residual network (ResNet) had shown remarkable performance in image recognition tasks, due to its shorter connections between layers close to the input and those close to the output. And densely connected convolutional network (DenseNet) had further improved recognition performance by dense feature reuses. To improve the performance of residual units in ResNet and feature reuses in DenseNet, we propose a simple and effective convolutional network architecture, named double reuses based residual network (DRRNet). DRRNet improves the residual unit of ResNet, where the feature reuse connections are added to combine all feature maps from the convolutional layers to produce the residual, and uses a residual reuse path outside units to reuse all residuals as the final feature maps for classification. Residual learning used in DRRNet can alleviate the vanishing-gradient problem. The double reuses including inner-unit feature reuses and outer-unit residual reuses effectively decrease computational cost as compared with dense connections in DenseNet, and further strengthen the forward feature propagation. DRRNet is evaluated on three object recognition benchmark datasets and an object detection dataset. In comparison with the state-of-the-art, DRRNet achieves a good balance between classification accuracy and computational cost, and optimal detection performance.