Feature Distance Research Articles

Simple exercise echocardiography parameters can enhance diagnosis of early heart failure with preserved ejection fraction (HFpEF)

Abstract Background Early heart failure with preserved ejection fraction (HFpEF) is characterised by exertional dyspnoea or intolerance. The non-specific nature of these symptoms and a lack of accurate diagnostic tools leave many undiagnosed and untreated. Whilst exercise right heart catheterisation is the recommended gold-standard, it is not widely available owing to technical challenges and its invasive nature. Exercise echocardiography offers a pragmatic compromise but its evidence-base in HFpEF diagnosis remains scarce, with significant variation in exercise protocols and assessment criteria. Purpose Using multimodality exercise imaging, the study aims to (1) correlate current recommended echocardiographic parameters at rest and exercise with functional capacity, symptom burden and adverse haemodynamic features (2) identify other echocardiographic parameters that interrogate mechanisms of exertional dyspnoea, thus improving early HFpEF diagnosis. Method Symptomatic patients with confirmed HFpEF or intermediate (INT) risk of HFpEF (as defined by the HFA-PEFF diagnostic algorithm) were prospectively enrolled alongside age-matched healthy control subjects. All participants underwent rest and 35W exercise echocardiography, exercise cardiac magnetic resonance imaging (e-CMR), 6-minute walk test (6MWT) and a Kansas City Cardiomyopathy Questionnaire (KCCQ-CSS). Results 46 HFpEF patients, 39 INT and 19 controls were recruited. The HFpEF and INT groups had similarly reduced functional capacity (6MWT HFpEF 363 vs. INT 384 vs. Control 492m) and worse patient-reported outcomes (KCCQ-CSS HFpEF 70 vs. INT 73 vs. Control 97) compared to age-matched controls. Despite more favourable serum NTproBNP, rest and stress E/e’ in the INT group, cardiac output augmentation during 35W exercise (ΔCI) on e-CMR was similarly blunted when compared to the HFpEF cohort (Table 1). Importantly, echocardiographic parameters in the current HFpEF diagnostic algorithm such as E/e’ and indexed left atrial volume (LAVI) do not correlate with 6MWT distance, KCCQ-CSS or ΔCI after adjusting for age, hypertension and body mass index (Figure A). Across this heterogeneous study cohort, rest and exercise LV septal S’ and RV S’ better correlated with 6MWT distance, KCCQ-CSS score and adverse haemodynamic features (ΔCI and right ventricular-pulmonary artery coupling) (Figures B-C). Conclusion Symptomatic patients at intermediate-risk of HFpEF share adverse haemodynamic features and functional impairment to those with diagnosed HFpEF. Using multimodality exercise imaging, we propose simple tissue doppler imaging parameters and assessment of cardiac reserve during exercise echocardiography better elicit mechanisms of exertional dyspnoea, beyond current recommendations. Further validation studies exploring these parameters are needed to advance the role of non-invasive imaging in early HFpEF diagnosis. Table 1 Figure 2

Multi-source Semi-supervised Condition Constrained Domain Adaptation for Process Fault Classification

Abstract To address inconsistent feature distributions caused by multiple working conditions in industrial processes and the lack of fault type labels, this paper proposes a Multi-source semi-supervised conditional constraint domain adaptation (MSSCCDA) fault classification method. Data from multiple working conditions are divided into multiple source domains and a target domain. A Convolutional neural network (CNN) extracts features from the multi-source domain data, and a triplet loss method reduces the feature distance between the same categories in different source domains. A semi-supervised conditional constraint domain adaptation method is proposed, which combines adversarial domain adaptation with a limited amount of labeled target domain data to pre-train the feature extractor. During adversarial training, center loss regularization is introduced as a conditional constraint for class feature alignment. The process of optimizing the feature extractor involves jointly leveraging adversarial loss to reduce inter-domain discrepancies between the source and target domains and minimizing class differences within target domain. Experiments conducted on the Tennessee-Eisman (TE) process and industrial three-phase flow (TFF) show that the average accuracy improves to 93% and 93%, respectively. The effectiveness of the proposed method is further validated through a comparison with two traditional domain adaptation approaches and five multi-source domain adaptation techniques.

Distance and Direction Matters: Risk Perception Among Residents Around a Dump Yard in Kerala, India.

Waste mismanagement is a growing concern in developing countries where unsustainable practices such as open dumping and open burning are rampant. This study examined the risk perceptions of the residents living in proximity to the Brahmapuram dump yard, situated in Ernakulam district of Kerala State, India- A site marked by persistent local protests, public outrage, and legal disputes arising from issues related to waste mismanagement. The study focused on the geospatial and sociodemographic factors that might influence these perceptions. A cross-sectional survey was conducted among 302 respondents living within 4 kilometers from the borders of the dump yard using a structured interview schedule. The responses of the participants were used to compute a risk perception score, which reflected participants' risk perception regarding the environment and their health. Among the participants in the study, those who lived within 2 kilometers (2.3 (95% CI 0.96, 3.7; p<0.001)), those who lived to the east (2.7 (95% CI 1.1, 4.2; p<0.001)) and those who reported perceiving strong malodor from the dump yard (2.0 (95% CI 0.54, 3.4; p=0.007)), had a higher risk perception in the multivariate linear regression model. Women had a lesser risk perception compared to men (-2.6 (95% CI -3.7, -1.4; p<0.001)). The findings highlight the importance of geospatial characteristics (distance and direction), malodor and gender differences in shaping the risk perceptions among the proximate residents living around a waste dump yard. Consideration of geospatial and sociodemographic determinants in risk assessment and management could potentially reduce the perceived risks and public discontent around waste management facilities.

Optimisation of an oxygen supply prediction model based on the K-nearest neighbours (KNN) algorithm using the Pearson correlation coefficient

Accurately predicting the oxygen supply in the basic oxygen furnace (BOF) steelmaking process is crucial for improving product quality, enhancing production efficiency and reducing costs. In this study, oxygen supply prediction models based on the K-nearest neighbours (KNN) algorithm were developed and compared using actual production data from the BOF. To further improve the accuracy of oxygen supply predictions for the BOF, this study optimises the algorithm using two methods: distance weighting and feature vector weighting based on the Pearson correlation coefficient. Experimental results indicate that the feature vector weighting method based on the Pearson correlation coefficient achieves better optimisation results compared to the distance weighting KNN algorithm. The prediction accuracy for oxygen supply reached 77.6% for an oxygen consumption error within ±200 Nm3, 89.8% for ±300 Nm3, and 95.7% for ±400 Nm3; additionally, this optimisation improved the endpoint hit rate for oxygen supply by 10.3% compared to the original algorithm. This model provides an effective reference for actual production and offers reliable insights for optimising other algorithms. Future research could further explore the application of other machine learning algorithms in oxygen supply prediction to achieve higher accuracy and broader applicability, thereby advancing the intelligent development of the BOF steelmaking process.

MDRN: Multi‐domain representation network for unsupervised domain generalization

AbstractIn deep neural networks, performance can degrade when test data distributions differ from training data. Unsupervised Domain Generalization (UDG) aims to improve generalization across unseen domains by leveraging multiple source domains without supervision. Traditional methods focus on extracting domain‐invariant features, potentially at the expense of feature space integrity and generalization potential. We presents a Multi‐Domain Representation Network (MDRN) for unsupervised multi‐domain learning. MDRN innovates by disentangling and preserving both domain‐invariant and domain‐specific features through an unsupervised cross‐domain reconstruction task. It employs content encoders for domain‐invariant features and multi‐domain style encoders for domain‐specific characteristics. By merging these features based on domain similarity, MDRN constructs a comprehensive feature space that enhances image reconstruction across domains. Additionally, MDRN integrates domain‐specific classifiers, which learn domain classification and provide weighted fusion of domain‐specific features. This design facilitates effective inter‐domain distance measurement and feature integration. Experiments on PACS and DomainNet show MDRN's superior performance over existing state‐of‐the‐art UDG approaches, highlighting its effectiveness in handling distribution shifts between source and target domains.

Machine learning materials properties with accurate predictions, uncertainty estimates, domain guidance, and persistent online accessibility

Abstract One compelling vision of the future of materials discovery and design involves the use of machine learning (ML) models to predict materials properties and then rapidly find materials tailored for specific applications. However, realizing this vision requires both providing detailed uncertainty quantification (model prediction errors and domain of applicability) and making models readily usable. At present, it is common practice in the community to assess ML model performance only in terms of prediction accuracy (e.g. mean absolute error), while neglecting detailed uncertainty quantification and robust model accessibility and usability. Here, we demonstrate a practical method for realizing both uncertainty and accessibility features with a large set of models. We develop random forest ML models for 33 materials properties spanning an array of data sources (computational and experimental) and property types (electrical, mechanical, thermodynamic, etc). All models have calibrated ensemble error bars to quantify prediction uncertainty and domain of applicability guidance enabled by kernel-density-estimate-based feature distance measures. All data and models are publicly hosted on the Garden-AI infrastructure, which provides an easy-to-use, persistent interface for model dissemination that permits models to be invoked with only a few lines of Python code. We demonstrate the power of this approach by using our models to conduct a fully ML-based materials discovery exercise to search for new stable, highly active perovskite oxide catalyst materials.

Analysis of Carbon Sink Benefits from Comprehensive Soil and Water Conservation in the Loess Hilly Gently Slope Aeolian Sand Region

Soil erosion has become an increasingly serious issue, drawing global attention. As one of the countries facing severe soil erosion in the world, China confronts significant ecological challenges. Against this backdrop, the country places great emphasis on soil conservation efforts, considering them a crucial component of ecological civilization construction. This study focuses on the carbon sink benefits of comprehensive soil conservation management in the loess hilly region and sandy slopes, using the Xiaonanshan Mountain small watershed in Youyu County, Shanxi Province, as a typical case for in-depth analysis. In terms of research methodology, an integrated monitoring approach combining fundamental data, measured data, and remote sensing data was developed. A comprehensive survey of the Xiaonanshan Mountain small watershed was conducted to categorize plant carbon pools and soil carbon pools, establish baseline scenarios, and utilize methods such as inverse distance spatial interpolation, sample calculation, and feature extraction to estimate forest carbon storage across different years and determine changes in soil and vegetation carbon storage. Simultaneously, data collection and preprocessing were carried out, including the gathering of fundamental data, field data collection, and internal data preprocessing. On this basis, a vegetation carbon storage model was constructed, and an assessment of soil carbon pool storage was conducted. The research results indicate that from 2002 to 2024, the continuous implementation of various soil conservation measures over 22 years has led to a significant increase in carbon storage within the Xiaonanshan Mountain small watershed. The vegetation carbon density of the entire small watershed increased from 14.66 t C/ha to 27.02 t C/ha, and the soil carbon density rose from 28.92 t C/ha to 32.48 t C/ha. The net carbon sink amount was 18,422.20 t C (corresponding to 67,548.08 t CO2e in terms of carbon dioxide equivalent). Populus simonii and Pinus sylvestris var. mongholica significantly contribute to the carbon sink; however, due to partial degradation of Populus simonii, its net carbon sink amount is less than that of Pinus sylvestris var. mongholica. Additionally, the carbon sink capacity of the small watershed exhibits spatial differences influenced by conservation measures, with high carbon density areas primarily concentrated within the range of Populus simonii, while low carbon density areas are mainly found in shrub zones. The increase in carbon storage within the small watershed is primarily attributed to the contributions of vegetation and soil carbon storage, indicating that comprehensive soil erosion management has a significant carbon accumulation effect; moreover, the annual growth rate of vegetation carbon storage exceeds that of soil carbon storage, with the proportion of soil carbon storage increasing year by year. Furthermore, the vegetation carbon sink, soil carbon sink, and total carbon sink of the small watershed were separately calculated. In terms of benefit analysis, the Xiaonanshan Mountain small watershed offers ecological benefits such as increased forest coverage, carbon fixation and oxygen release, and biodiversity conservation; from an economic perspective, the value of carbon trading is substantial, promoting soil conservation and rural revitalization, with the total value of timber reaching 7.6 million yuan, of which the value of standing timber constitutes the largest proportion; social benefits include the improvement of environmental landscapes, stimulation of ecological tourism, and attraction of investment, with the Xiaonanshan Mountain Ecological Park receiving numerous visitors and generating significant tourism revenue. This research provides a theoretical basis and data foundation for comprehensive soil conservation management in project areas or small watersheds within the loess hilly and sandy slope regions, offering technical and methodological support for other soil conservation carbon sink projects in the area.

Lane extraction from trajectories at road intersections based on Graph Transformer Network

Lane-level road networks are crucial components of high-precision maps and play a significant role in intelligent transportation systems. Extracting lane-level road networks at intersections presents considerable challenges due to the complex structures of intersections and diverse driving behaviors. A graph-learning based method is proposed for extracting lanes from high-precision trajectories at road intersections. A trajectory relation graph is designed to encode the directional, shape, and distance features of trajectories, capturing both the intrinsic and extrinsic relationships between trajectories. Subsequently, a Graph Transformer Network is developed to extract a representative subset of trajectories as lanes. To alleviate the problem of generating missing and extraneous lanes, a set-based lane extraction loss is introduced to achieve implicit pruning of redundancy through the attention mechanism. Comprehensive experimental results demonstrate that the proposed method outperforms state-of-the-art methods in three positional and topological accuracy metrics. The method achieves lane extraction with minimal omissions and redundancies and exhibits strong performance in complex scenarios such as U-turns, lane merging, and lane diverging regions.

A class alignment network based on self-attention for cross-subject EEG classification.

Due to the inherent variability in EEG signals across different individuals, domain adaptation and adversarial learning strategies are being progressively utilized to develop subject-specific classification models by leveraging data from other subjects. These approaches primarily focus on domain alignment and tend to overlook the critical task-specific class boundaries. This oversight can result in weak correlation between the extracted features and categories. To address these challenges, we propose a novel model that uses the known information from multiple subjects to bolster EEG classification for an individual subject through adversarial learning strategies. Our method begins by extracting both shallow and attention-driven deep features from EEG signals. Subsequently, we employ a class discriminator to encourage the same-class features from different domains to converge while ensuring that the different-class features diverge. This is achieved using our proposed discrimination loss function, which is designed to minimize the feature distance for samples of the same class across different domains while maximizing it for those from different classes. Additionally, our model incorporates two parallel classifiers that are harmonious yet distinct and jointly contribute to decision-making. Extensive testing on two publicly available EEG datasets has validated our model's efficacy and superiority.

Self-supervised Learning Approach for Anomaly Detection in Rotating Machinery

Early fault detection in rotating machinery needs careful expert analysis of vibration data for monitoring a component state. Online methods that automatically set a threshold and raise an alarm when the vibration signature is anomalous are meant to efficiently manage key assets in a preventive maintenance plan. In recent years a focus has raised on data driven methods in parallel with the increasing attention towards machine learning and, particularly, deep learning models. In this regard, for rotating equipment components, an important aspect relates to labelled data scarcity for supervised training. On the other hand, the advent of the Internet of Things allows to gather data from multiple assets with relevant information on the asset state itself. Self-supervised learning methods in deep learning application are currently tackling this problem. Investigating Self-learning approaches to integrate domain knowledge and learn relevant features from unlabeled data is therefore important for condition monitoring applications. In this paper a methodology is proposed based on cycle consistency representation learning for training an embedder network on univariate unlabeled data. In order to learn a distance metric in the embedding space the original data are transformed to generate sequences of augmented inputs to enforce learnable pattern similarity in the augmented pairs. A differentiable cycle-consistency loss is chosen to maximize the numbers of augmented pairs in the learned embedding space that have minimum features distance. The pretext task in the described self-supervised setting aims to train a feature extractor for discriminating dissimilar samples in the embedding space by a distance metric and to provide a useful representation for down-stream tasks. The paper analyzes the performance of the approach for anomaly detection in rotating machinery. The methodology is tested on vibration data provided by the Center for Intelligent Maintenance Systems (IMS), University of Cincinnati, considering different accelerated life test campaigns. The data were collected to monitor the fault development in bearings and the model shows how the learned embedding space discriminates effectively anomalous samples from normal ones in the degradation stages of the bearings.

How the Galaxy–Halo Connection Depends on Large-scale Environment

We investigate the connection between galaxies, dark matter halos, and their large-scale environments at z = 0 with Illustris TNG300 hydrodynamic simulation data. We predict stellar masses from subhalo properties to test two types of machine learning (ML) models: explainable boosting machines (EBMs) with simple galaxy environment features and E(3) -invariant graph neural networks (GNNs). The best-performing EBM models leverage spherically averaged overdensity features on 3 Mpc scales. Interpretations via SHapley Additive exPlanations also suggest that in the context of the TNG300 galaxy–halo connection, simple spherical overdensity on ∼3 Mpc scales is more important than cosmic web distance features measured using the DisPerSE algorithm. Meanwhile, a GNN with connectivity defined by a fixed linking length, L, outperforms the EBM models by a significant margin. As we increase the linking length scale, GNNs learn important environmental contributions up to the largest scales we probe (L = 10 Mpc). We conclude that 3 Mpc distance scales are most critical for describing the TNG galaxy–halo connection using the spherical overdensity parameterization, but that information on larger scales, which is not captured by simple environmental parameters or cosmic web features, can further augment these models. Our study highlights the benefits of using interpretable ML algorithms to explain models of astrophysical phenomena, and the power of using GNNs to flexibly learn complex relationships directly from data while imposing constraints from physical symmetries.

Image shadow removal based on pseudo-illuminated region and log domain transformation

ABSTRACT The presence of shadows in images can significantly degrade image quality. To address the challenges, this study utilizes pseudo-illumination regions and logarithmic domain transformations. Firstly, the input shadow image and its corresponding shadow region are segmented and clustered into multiple image blocks. Then, a comprehensive consideration of distance, color, and texture features is employed to construct pseudo illumination regions. Subsequently, the designed logarithmic domain transformation function is utilized for shadow removal. Finally, color correction and optimization of shadow edges complete the overall process. The feasibility of this approach has been preliminarily verified through experimental analysis.

A Clustering Approach to Unveil User Similarities in 6-DoF Extended Reality Applications

The advent in our daily life of Extended Reality (XR) technologies, such as Virtual and Augmented Reality, has led to the rise of user-centric systems, offering higher level of interaction and presence in virtual environments. In this context, understanding the actual interactivity of users is still an open challenge and a key step to enabling user-centric system. In this work, our goal is to construct an efficient clustering tool for 6 Degree-of-Freedom (DoF) navigation trajectories by extending the applicability of existing behavioural tool. Specifically, we first compare the navigation in 6-DoF with its 3-DoF counterpart, highlighting the main differences and novelties. Then, we investigate new metrics aimed at better modelling behavioural similarities between users in a 6-DoF system. More concretely, we define and compare 11 similarity metrics which are based on different distance features ( i.e. , user positions in the 3D space, user viewing directions) and distance measurements ( i.e. , Euclidean, Geodesic, angular distance). Our solutions are validated and tested on real navigation paths of users interacting with dynamic volumetric media in both 6-DoF Virtual Reality and Augmented Reality conditions. Results show that metrics based on both user position and viewing direction better perform in detecting user similarity while navigating in a 6-DoF system. Such easy-to-use but robust metrics allow us to answer a fundamental question for user-centric systems: “how do we detect if users look at the same content in 6-DoF?”, opening the gate to new solutions based on users interactivity, such as viewport prediction, live streaming services optimised based on users behaviour but also for user-based quality assessment methods.

Research on a Multi-Scale Clustering Method for Buildings Taking into Account Visual Cognition

Building clustering is a key problem that needs to be solved in the realization of the automatic synthesis of large-scale maps. The selection of different feature and spatial distance calculation methods has a great impact on the clustering results, and the need to manually select appropriate feature and distance metrics leads to the problem of not being able to fully consider the complexity and diversity of buildings. In this paper, we propose a multi-scale clustering method for buildings that takes visual perception into account using the Gestalt principle to simulate how humans classify buildings through visual perception. Moreover, by analyzing the spatial features and texture attributes of buildings, a visual distance is designed to be used as a condition for building classification to assess the similarity between buildings, solving the complexity of manually selecting feature vectors and spatial distances and realizing the adaptive selection of features. Through experimental validation at different scales (macro, meso and micro), the present method is able to achieve the accurate clustering of buildings, and a frequency threshold of 91% is found, which is able to determine the optimal clustering results. The experimental results show that the proposed method can not only fully consider the complexity and diversity of buildings but also effectively support the understanding and analysis of urban spatial structure and provide a scientific decision-making basis for urban planning and management.

Bearing fault diagnosis based on enhanced Canberra distance feature in SDP image

Abstract Feature enhancement is important in mechanical equipment fault diagnosis. A limited set of characteristic parameters is insufficient for diagnosing bearing signals with multiple fault types. The presence of noise increases the difficulty of extracting fault features from images. To address the challenge of diagnosing rolling bearing faults in complex environments, this study presents an enhanced weighted image fusion framework aimed at enhancing fault features within the images, which enables accurate diagnosis of bearing faults using a limited number of features. The proposed method encompasses four distinct stages. In the first stage, a symmetrized dot pattern method is employed to transform one-dimensional time-series data into two-dimensional images, visualizing the signal in a 2D format. In the second stage, image binarization and an improved weighted fusion method are utilized to simplify subsequent processing and enhance the image features. The third stage involves extracting the image’s contrast and maximum singular value to improve the Canberra distance calculation. Finally, the enhanced Canberra distance is used for classifying bearing faults. Performance testing of the image feature enhancement is conducted on various datasets containing rolling bearings. Comparative experiments with alternative enhancement methods demonstrate the superiority of the proposed improved weighted image fusion framework. Comparative experiments with the original Canberra distance validate the effectiveness of the enhanced Canberra distance. Additionally, experiments conducted in noisy environments confirm the robustness of the proposed approach. Furthermore, the image feature enhancement method is applied to other bearing datasets, and the experimental results demonstrate its effectiveness in enhancing fault feature representation and achieving accurate diagnosis of rolling bearings.

Stance Detection in the Context of Fake News—A New Approach

Online social networks (OSNs) are inundated with an enormous daily influx of news shared by users worldwide. Information can originate from any OSN user and quickly spread, making the task of fact-checking news both time-consuming and resource-intensive. To address this challenge, researchers are exploring machine learning techniques to automate fake news detection. This paper specifically focuses on detecting the stance of content producers—whether they support or oppose the subject of the content. Our study aims to develop and evaluate advanced text-mining models that leverage pre-trained language models enhanced with meta features derived from headlines and article bodies. We sought to determine whether incorporating the cosine distance feature could improve model prediction accuracy. After analyzing and assessing several previous competition entries, we identified three key tasks for achieving high accuracy: (1) a multi-stage approach that integrates classical and neural network classifiers, (2) the extraction of additional text-based meta features from headline and article body columns, and (3) the utilization of recent pre-trained embeddings and transformer models.

Ensemble Learning Based on Feature Selection and Distance Normalization for Enhancing Corn and Weed Classification

Ensemble Learning Based on Feature Selection and Distance Normalization for Enhancing Corn and Weed Classification

Parasite β-diversity along a stream: effect of distance and environment

Parasites can provide suitable models for studying β-diversity due to their strict dependence on both the environment and the biology and distribution of their hosts, aiding in the interpretation of any patterns that hosts can display. With the aim of quantifying the relative importance of host features, environmental factors and spatial distances as drivers of fish parasite β-diversity along a unidirectional gradient, the structure of fish parasite assemblages was analysed using generalised dissimilarity models (GDMs). A total of 150 poeciliid fish were examined for larval trematodes, recording host features and physical parameters of each sampling site along the stream. Differences among digenean communities increased when Strahler order changed along the stream, associated with increasing species richness and abundance downstream. Environmental gradient, spatial distance and host features were identified as significant determining factors of species turnover, with conductivity being the most important, followed by spatial distance. In the present study, environmental variables were spatially structured along the stream, their effects as structurers of parasite β-diversity being higher than the pure environmental or the pure distance effect. Such predominance prevents us from establishing at what point on the continuum from niche to neutrality these communities are located. Results from the present research contribute to improving our knowledge of the factors that shape parasite community changes, and underline the importance of considering the pure and shared effects of spatial, environmental and host feature factors in order to determine the real contribution of each one as a determinant of parasite β- diversity.

Modeling and Optimization of Interaction Behavior between Right Turning Motor Vehicles and Pedestrians at Signalized Intersections based on Machine Learning

This project first conducted a comprehensive analysis of the interaction process between pedestrians and right turning motor vehicles and its influencing factors through a summary of existing literature and observation of actual situations. On the basis of comprehensive and accurate data, machine learning algorithms were used to model and predict the interaction behavior between pedestrians and right turning motor vehicles. By comparing different algorithms, the optimal models for predicting the yielding behavior of right turning motor vehicles and pedestrian crossing behavior in the current dataset were obtained. According to the model evaluation report, the obtained models have better predictive performance on the test set. And based on the obtained optimal model, the feature importance ranking was carried out for the pedestrian crossing model and the right turn motor vehicle yielding model, and the importance scores of the influencing factors of pedestrian crossing behavior and right turn motor vehicle yielding behavior were obtained respectively. Finally, a comparison was made between the modeling effects of the crossing gap feature group and the speed distance feature group, further clarifying the importance of the speed distance index for right turning motor vehicles in predicting interactive behavior.

Relevance Pooling Guidance and Class-Balanced Feature Enhancement for Fine-Grained Oriented Object Detection in Remote Sensing Images

Fine-grained object detection in remote sensing images is highly challenging due to class imbalance and high inter-class indistinguishability. The strategies employed by most existing methods to resolve these two challenges are relatively rudimentary, resulting in suboptimal model performance. To address these issues, we propose a fine-grained object-oriented detection method based on relevance pooling guidance and class balance feature enhancement. Firstly, we propose a global attention mechanism that dynamically retains spatial features pertinent to objects through relevance pooling during down-sampling, thereby enabling the model to acquire more discriminative features. Next, a class balance correction module is proposed to alleviate the class imbalance problem. This module employs feature translation and a learnable reinforcement coefficient to highlight the boundaries of tail class features while maintaining their distinctiveness. Furthermore, we present an enhanced contrastive learning strategy. By dynamically adjusting the contribution of inter-class samples and intra-class similarity measures, this strategy not only constrains inter-class feature distances but also facilitates tighter intra-class clustering, making it more suitable for imbalanced datasets. Evaluation on the FAIR1M and MAR20 datasets demonstrates that our method is superior compared to other methods in object detection and achieves 46.44 and 85.05% mean average precision, respectively.