Sparse Feature Research Articles

Poisson phase diversity algorithm with automatic registration for sparse fluorescent images

The phase diversity technique is a powerful tool for aberration detection and high-resolution image restoration. This is achieved by employing a pair of images, one captured in focus and the other with a specific degree of defocus. It presents a practical alternative or complementary method to adaptive optics. This paper introduces a modified phase diversity algorithm based on the Poisson noise model for biological fluorescence imaging with sparse characteristics. The proposed method has demonstrated high-precision aberration estimation and high-resolution image reconstruction. Additionally, an image registration method is proposed for image registration between the focused and defocused images. The simulation analysis and verification of the algorithm were conducted using images of microbeads and HeLa cells. For experimental purposes, actual captured fluorescence microbead images and blurred zebrafish images due to aberrations were selected. Both simulation and experimental results indicate that the method is effective and accurate.

Multiscale local sparsity and prior learning algorithm for Cherenkov-excited luminescence scanned tomography reconstruction

Cherenkov-excited luminescence scanned tomography (CELST) is an emerging imaging technique and its potential applications during radiation therapy have just recently been explored. The aim of CELST is to recover the distribution of luminescent probes from emission photons. However, CELST images tend to suffer from low resolution and degraded image quality due to light multiple scattering and limited boundary measurements. Therefore, inaccurate information about the status of the luminescent probe is provided. To accurately capture the sparsity characterization of a luminescent probe and achieve the high-quality image, a novel reconstruction method, to our knowledge, is proposed for CELST by combining a sparse prior with an attention network, termed LKSVD-Net. A multiscale learned KSVD is first incorporated to obtain the local sparsity information of a luminescent probe. Subsequently, a prior attention network is designed to leverage the prior features related to the measurements. The multiscale sparsity and prior features are finally combined to complete the image reconstruction. Experimental results demonstrate that the LKSVD-Net can notably enhance image quality even in a 20 dB signal-to-noise ratio (SNR). Furthermore, the proposed LKSVD-Net yields improved quantitative accuracy for 4 mm diameter probes with an edge-to-edge distance of 2 mm. The results demonstrate that LKSVD-Net improves the peak signal-to-noise ratio (PSNR) by approximately 15.1%, structural similarity index measure (SSIM) by about 95.8%, and Pearson correlation (PC) by around 3% compared to Tikhonov regularization.

Multiscale Dynamic Attention and Hierarchical Spatial Aggregation for Few-Shot Object Detection

Few-shot object detection (FSOD) remains a critical challenge in computer vision, where the limited training data significantly hinder model performance. Existing methods suffer from poor robustness and accuracy, primarily due to scale sparsity and inadequate feature extraction. In this paper, we propose MDA-HAPP, a novel framework built on a transfer learning architecture and a two-stage object detection approach, specifically designed to address these issues. The key innovations of MDA-HAPP include 1. MultiScale-DynaAttention, a novel attention module that enhances feature extraction by integrating multi scale convolutions into channel attention and applying a dynamic pooling ratio to spatial attention, with residual connections to improve robustness; 2. hierarchical adaptive-pyramid pooling, designed based on a spatial pyramid pooling (SPP) structure, extracts multiscale features from intermediate layers and dynamically adjusts pooling strategies. These features are then fed into a dual-branch detection head for comprehensive results.The experimental results on the PASCAL VOC and COCO datasets show that MDA-HAPP achieves significant improvements across different K-shot settings. Specifically, the model demonstrates an up to 9.8% gain in AP75 on PASCAL VOC for K-shot values of 10 and an up to 3.7% improvement on COCO for K-shot values of 30. These results confirm its superior performance in FSOD and highlight its potential for real-world applications.

Analysis of Sparse Trajectory Features Based on Mobile Device Location for User Group Classification Using Gaussian Mixture Model

Location data collected from mobile devices via global positioning system often lack semantic information and can form sparse trajectories in space and time. This study investigates whether user age groups can be accurately classified solely from such sparse spatial–temporal trajectories. We propose a feature extraction method based on a Gaussian mixture model (GMM), which assigns representative points (RPs) by clustering the location data and aggregating user trajectories into these RPs. We then construct three machine learning (ML) models—support vector classifier (SVC), random forest (RF), and deep neural network (DNN)—using the GMM-based features and compare their performance with that of the improved DNN (IDNN), which is an existing feature extraction approach. In our experiments, we introduced a missing value ratio θth to quantify trajectory sparsity and analyzed the effect of trajectory sparsity on the classification accuracy and generalizability performance of the ML models. The results indicate that GMM-based features outperform IDNN-based features in both classification accuracy and generalization performance. Notably, the RF model achieved the highest accuracy, whereas the SVC model displayed stable generalizability. As the missing value ratio θth increases, the IDNN becomes more susceptible to overfitting, whereas the GMM-based approach preserves accuracy and robustness. These findings suggest that sparse trajectories can still offer meaningful classification performance with appropriate feature design and model selection even without semantic information. This approach holds promise for domains where large-scale, sparse trajectory data are common, including urban planning, marketing analysis, and public policy.

Side-Scan Sonar Small Objects Detection Based on Improved YOLOv11

Underwater object detection using side-scan sonar (SSS) remains a significant challenge in marine exploration, especially for small objects. Conventional methods for small object detection face various obstacles, such as difficulties in feature extraction and the considerable impact of noise on detection accuracy. To address these issues, this study proposes an improved YOLOv11 network named YOLOv11-SDC. Specifically, a new Sparse Feature (SF) module is proposed, replacing the Spatial Pyramid Pooling Fast (SPPF) module from the original YOLOv11 architecture to enhance object feature selection. Furthermore, the proposed YOLOv11-SDC integrates a Dilated Reparam Block (DRB) with a C3k2 module to broaden the model’s receptive field. A Content-Guided Attention Fusion (CGAF) module is also incorporated prior to the detection module to assign appropriate weights to various feature maps, thereby emphasizing the relevant object information. Experimental results clearly demonstrate the superiority of YOLOv11-SDC over several iterations of YOLO versions in detection performance. The proposed method was validated through extensive real-world experiments, yielding a precision of 0.934, recall of 0.698, mAP@0.5 of 0.825, and mAP@0.5:0.95 of 0.598. In conclusion, the improved YOLOv11-SDC offers a promising solution for detecting small objects in SSS images, showing substantial potential for marine applications.

Array Three-Dimensional SAR Imaging via Composite Low-Rank and Sparse Prior

Array three-dimensional (3D) synthetic aperture radar (SAR) imaging has been used for 3D modeling of urban buildings and diagnosis of target scattering characteristics, and represents one of the significant directions in SAR development in recent years. However, sparse driven 3D imaging methods usually only capture the sparse features of the imaging scene, which can result in the loss of the structural information of the target and cause bias effects, affecting the imaging quality. To address this issue, we propose a novel array 3D SAR imaging method based on composite sparse and low-rank prior (SLRP), which can achieve high-quality imaging even with limited observation data. Firstly, an imaging optimization model based on composite SLRP is established, which captures both sparse and low-rank features simultaneously by combining non-convex regularization functions and improved nuclear norm (INN), reducing bias effects during the imaging process and improving imaging accuracy. Then, the framework that integrates variable splitting and alternative minimization (VSAM) is presented to solve the imaging optimization problem, which is suitable for high-dimensional imaging scenes. Finally, the performance of the method is validated through extensive simulation and real data experiments. The results indicate that the proposed method can significantly improve imaging quality with limited observational data.

Compositional transformations can reasonably introduce phenotype-associated values into sparse features.

It was recently argued 1 that an analysis of tumor-associated microbiome data 2 is invalid because features that were originally very sparse (genera with mostly zero read counts) became associated with the phenotype following batch correction 1 . Here, we examine whether such an observation should necessarily indicate issues with processing or machine learning pipelines. We show counterexamples using the centered log ratio (CLR) transformation, which is often used for analysis of compositional microbiome data 3 . The CLR transformation has similarities to Voom-SNM 4,5 , the batch-correction method brought into question 1,2 , yet is a sample-wise operation that cannot, in itself, "leak" information or invalidate downstream analyses. We show that because the CLR transformation divides each value by the geometric mean of its sample, common imputation strategies for missing or zero values result in transformed features that are associated with the geometric mean. Through analyses of both synthetic and vaginal microbiome datasets we demonstrate that when the geometric mean is associated with a phenotype, sparse and CLR-transformed features will also become associated with it. We re-analyze features highlighted by Gihawi et al. 1 and demonstrate that the phenomenon of sparse features becoming phenotype-associated can also be observed after a CLR transformation, which serves as a counterexample to the claim that such an observation necessarily means information leakage. While we do not intend to address other concerns regarding tumor microbiome analyses 1,6 , validate Poore et al.'s 2 results, or evaluate batch-correction pipelines, we conclude that because phenotype-associated features that were initially sparse can be created by a sample-wise transformation that cannot artifactually inflate machine learning performance, their detection is not independently sufficient to demonstrate information leakage in machine learning pipelines. Microbiome data is multivariate, and as such, a value of zero carries a different meaning for each sample. Many transformations, including CLR and other batch-correction methods, are likewise multivariate, and, as these issues demonstrate, each individual feature should be interpreted with caution.

Sparse multi-label feature selection via pseudo-label learning and dynamic graph constraints

Sparse multi-label feature selection via pseudo-label learning and dynamic graph constraints

Over-the-Air Fusion of Sparse Spatial Features for Integrated Sensing and Edge AI over Broadband Channels

Over-the-Air Fusion of Sparse Spatial Features for Integrated Sensing and Edge AI over Broadband Channels

Feature selection through adaptive sparse learning for scene recognition

Scene recognition is an important and challenging task in the field of computer vision. Current research typically focuses on local features in scene images by utilizing pretrained convolutional neural networks (CNN) to obtain a feature dictionary. However, these local features often contain similar features to other scene categories, leading to feature confusion and subsequently low accuracy in scene recognition. Therefore, focusing on local features for scene recognition is controversial. In contrast to existing works, we consider extracting common features within the same category from scene images and using them as discriminative features between different categories. We propose a scene recognition method based on adaptive sparse learning feature selection. This method leverages sparse learning to distinguish the contribution of each feature in the deep features to the classification task, aiming to construct a salient feature dictionary that describes the importance of features in scene images. Subsequently, an adaptive weight optimization method is employed through alternating updates to automatically adjust feature weights, enabling the selection of common features within the same scene category from the compact dictionary. Experimental results on multiple benchmark scene recognition datasets demonstrate that our proposed method is superior to state-of-the-art methods.

$\ell_1$DecNet+: A New Architecture Framework by $ℓ_1$ Decomposition and Iteration Unfolding for Sparse Feature Segmentation

$\ell_1$DecNet+: A New Architecture Framework by $ℓ_1$ Decomposition and Iteration Unfolding for Sparse Feature Segmentation

Revisiting non-learned operators based deep learning for image classification: a lightweight directional-aware network

Due to the stable feature representation capability provided by non-learned operators, the integration with deep learning models, i.e., non-learned operator based deep learning models, has become a paradigm, however, performance-wise, it is still not promising. In this paper, by revisiting non-learned operator based deep learning models, we reveal the reasons for their underperformance: lack of geometric invariance, insufficient sparsity, and neglect of directional importance. In response, we present a Lightweight Directional-Aware Network (LDAN) for image classification. Specifically, to generate sparse geometric-invariant features, we propose a ShearletNet to capture multi-directional features in three different levels. Then, a Directional-Aware module is designed to highlight the discriminative multi-directional features and generate multi-scale features. Finally, a Pointwise Convolution module is used to integrate the multi-directional features with the multi-scale ones for reducing the computational resources. Experiments on the commonly used CIFAR10, CIFAR100, Self-Taught Learning 10 (STL10), and Tiny ImageNet datasets demonstrate the efficiency and effectiveness of the proposed LDAN. Compared to the existing non-learned operator based models, LDAN reduces the parameter count by 80.83% while achieving a 6.32% increase in accuracy.

A Low-Cost 3D Mapping System for Indoor Scenes Based on 2D LiDAR and Monocular Cameras

The cost of indoor mapping methods based on three-dimensional (3D) LiDAR can be relatively high, and they lack environmental color information, thereby limiting their application scenarios. This study presents an innovative, low-cost, omnidirectional 3D color LiDAR mapping system for indoor environments. The system consists of two two-dimensional (2D) LiDARs, six monocular cameras, and a servo motor. The point clouds are fused with imagery using a pixel-spatial dual-constrained depth gradient adaptive regularization (PS-DGAR) algorithm to produce dense 3D color point clouds. During fusion, the point cloud is reconstructed inversely based on the predicted pixel depth values, compensating for areas of sparse spatial features. For indoor scene reconstruction, a globally consistent alignment algorithm based on particle filter and iterative closest point (PF-ICP) is proposed, which incorporates adjacent frame registration and global pose optimization to reduce mapping errors. Experimental results demonstrate that the proposed density enhancement method achieves an average error of 1.5 cm, significantly improving the density and geometric integrity of sparse point clouds. The registration algorithm achieves a root mean square error (RMSE) of 0.0217 and a runtime of less than 4 s, both of which outperform traditional iterative closest point (ICP) variants. Furthermore, the proposed low-cost omnidirectional 3D color LiDAR mapping system demonstrates superior measurement accuracy in indoor environments.

Multipath Mitigation in Single-Frequency Multi-GNSS Tightly Combined Positioning via a Modified Multipath Hemispherical Map Method

Multipath is a source of error that limits the Global Navigation Satellite System (GNSS) positioning precision in short baselines. The tightly combined model between systems increases the number of observations and enhances the strength of the mathematical model owing to the continuous improvement in GNSS. Multipath mitigation of the multi-GNSS tightly combined model can improve the positioning precision in complex environments. Interoperability of the multipath hemispherical map (MHM) models of different systems can enhance the performance of the MHM model due to the small multipath differences in single overlapping frequencies. The adoption of advanced sidereal filtering (ASF) to model the multipath for each satellite brings computational challenges owing to the characteristics of the multi-constellation heterogeneity of different systems; the balance efficiency and precision become the key issues affecting the performance of the MHM model owing to the sparse characteristics of the satellite distribution. Therefore, we propose a modified MHM method to mitigate the multipath for single-frequency multi-GNSS tightly combined positioning. The method divides the hemispherical map into 36 × 9 grids at 10° × 10° resolution and then searches with the elevation angle and azimuth angle as independent variables to obtain the multipath value of the nearest point. We used the k-d tree to improve the search efficiency without affecting precision. Experiments show that the proposed method improves the mean precision over ASF by 10.20%, 10.77%, and 9.29% for GPS, BDS, and Galileo satellite single-difference residuals, respectively. The precision improvements of the modified MHM in the E, N, and U directions were 32.82%, 40.65%, and 31.97%, respectively. The modified MHM exhibits greater performance and behaves more consistently.

Daily Engine Performance Trending Using Common Flight Regime Identification

Abstract Accurate flight regime identification is critical for enhancing aircraft efficiency and safety. Traditionally, predictive models for aircraft operation have relied on complex, black-box machine learning techniques that lack transparency. This study introduces a more interpretable approach by leveraging the New Comprehensive Modular Aero-Propulsion System Simulation (N-CMAPSS) dataset and combining expanding window classification, voting schemes, and spectral clustering to detect distinct flight regimes. The method applies elastic registration to align time-shifted patterns and functional principal component analysis (FPCA) to reduce dimensionality, capturing core dynamics across flight regimes. These transformed features are fed into a genetic algorithm (GA)-assisted orthogonal matching pursuit (OMP) for sparse feature selection. Through evolutionary selection, crossover, and mutation, the most informative features are identified, enabling accurate predictions while maintaining transparency. This method outperforms more complex models in certain test cases, offering a balance between accuracy and interpretability that is essential for predictive maintenance and safety applications.

Automated derivation of diagnostic criteria for lung cancer using natural language processing on electronic health records: a pilot study

BackgroundThe digitisation of healthcare records has generated vast amounts of unstructured data, presenting opportunities for improvements in disease diagnosis when clinical coding falls short, such as in the recording of patient symptoms. This study presents an approach using natural language processing to extract clinical concepts from free-text which are used to automatically form diagnostic criteria for lung cancer from unstructured secondary-care data.MethodsPatients aged 40 and above who underwent a chest x-ray (CXR) between 2016 and 2022 were included. ICD-10 and unstructured data were pulled from their electronic health records (EHRs) over the preceding 12 months to the CXR. The unstructured data were processed using named entity recognition to extract symptoms, which were mapped to SNOMED-CT codes. Subsumption of features up the SNOMED-CT hierarchy was used to mitigate against sparse features and a frequency-based criteria, combined with univariate logarithmic probabilities, was applied to select candidate features to take forward to the model development phase. A genetic algorithm was employed to identify the most discriminating features to form the diagnostic criteria.Results75002 patients were included, with 1012 lung cancer diagnoses made within 12 months of the CXR. The best-performing model achieved an AUROC of 0.72. Results showed that an existing ‘disorder of the lung’, such as pneumonia, and a ‘cough’ increased the probability of a lung cancer diagnosis. ‘Anomalies of great vessel’, ‘disorder of the retroperitoneal compartment’ and ‘context-dependent findings’, such as pain, statistically reduced the risk of lung cancer, making other diagnoses more likely. The performance of the developed model was compared to the existing cancer risk scores, demonstrating superior performance.ConclusionsThe proposed methods demonstrated success in leveraging unstructured secondary-care data to derive diagnostic criteria for lung cancer, outperforming existing risk tools. These advancements show potential for enhancing patient care and results. However, it is essential to tackle specific limitations by integrating primary care data to ensure a more thorough and unbiased development of diagnostic criteria. Moreover, the study highlights the importance of contextualising SNOMED-CT concepts into meaningful terminology that resonates with clinicians, facilitating a clearer and more tangible understanding of the criteria applied.

A new approach to biometric wood log traceability combining traditional methods and deep learning

This paper focuses on the biometric traceability of oak logs using images of cross-sections. The images were acquired in two temporally separated image acquisition sessions, and we want to find the correct matches for the images of the second session from the pool of images of the first session. No biometric traceability method has yet been proposed for oak logs and they differ greatly from softwood logs. In this context, we present a new method consisting of two steps. The first one involves extracting visible features from log-end images using the MODIFScale-Invariant Feature Transform (SIFT) method and the SuperPoint architecture. Then, in the second step, the extracted features are matched to verify whether two images correspond to the same log. For this, we consider the deep neural network LightGlue that is well-known for efficiently matching sparse local features between pairs of images. This new approach was compared with two recent state-of-the-art methods, including a significant evolution of one of them. The experiments were carried out on two datasets, including a new and large dataset of almost 25k images. The results show the performance of the new method for identifying oak logs, significantly outperforming the most recent ones. Source code and pre-trained models are available at https://github.com/Braquemarok/ATECH2024, while the image database is accessible at https://doi.org/10.57745/9DBCL4.1

Video-based Soft Tissue Deformation Tracking for Laparoscopic Augmented Reality-based Navigation in Kidney Surgery.

Minimally invasive surgery (MIS) remains technically demanding due to the difficulty of tracking hidden critical structures within the moving anatomy of the patient. In this study, we propose a soft tissue deformation tracking augmented reality (AR) navigation pipeline for laparoscopic surgery of the kidneys. The proposed navigation pipeline addresses two main sub-problems: the initial registration and deformation tracking. Our method utilizes preoperative MR or CT data and binocular laparoscopes without any additional interventional hardware. The initial registration is resolved through a probabilistic rigid registration algorithm and elastic compensation based on dense point cloud reconstruction. For deformation tracking, the sparse feature point displacement vector field continuously provides temporal boundary conditions for the biomechanical model. To enhance the accuracy of the displacement vector field, a novel feature points selection strategy based on deep learning is proposed. Moreover, an ex-vivo experimental method for internal structures error assessment is presented. The ex-vivo experiments indicate an external surface reprojection error of 4.07 ± 2.17mm and a maximum mean absolutely error for internal structures of 2.98mm. In-vivo experiments indicate mean absolutely error of 3.28 ± 0.40mm and 1.90±0.24mm, respectively. The combined qualitative and quantitative findings indicated the potential of our AR-assisted navigation system in improving the clinical application of laparoscopic kidney surgery.

Sparse Spike Feature Learning to Recognize Traceable Interictal Epileptiform Spikes.

Interictal epileptiform spikes (spikes) and epileptogenic focus are strongly correlated. However, partial spikes are insensitive to epileptogenic focus, which restricts epilepsy neurosurgery. Therefore, identifying spike subtypes that are strongly associated with epileptogenic focus (traceable spikes) could facilitate their use as reliable signal sources for accurately tracing epileptogenic focus. However, the sparse firing phenomenon in the transmission of intracranial neuronal discharges leads to differences within spikes that cannot be observed visually. Therefore, neuro-electro-physiologists are unable to identify traceable spikes that could accurately locate epileptogenic focus. Herein, we propose a novel sparse spike feature learning method to recognize traceable spikes and extract discrimination information related to epileptogenic focus. First, a multilevel eigensystem feature representation was determined based on a multilevel feature representation module to express the intrinsic properties of a spike. Second, the sparse feature learning module expressed the sparse spike multi-domain context feature representation to extract sparse spike feature representations. Among them, a sparse spike encoding strategy was implemented to effectively simulate the sparse firing phenomenon for the accurate encoding of the activity of intracranial neurosources. The sensitivity of the proposed method was 97.1%, demonstrating its effectiveness and significant efficiency relative to other state-of-the-art methods.

A novel structure adaptive new information priority grey Bernoulli model and its application in China's renewable energy production

At present, the global energy structure is undergoing major changes. China is in the transition period of energy structure. Accurately anticipating future energy trends is critical for China's energy structure and modernization. Considering the uncertainty and sparsity characteristics of China's energy system sequence, this study examines China's renewable energy scenario using the grey model with limited sample and uncertain system modelling features. Renewable energy is affected by a variety of uncertain factors, exhibiting a range of complicated traits including nonlinearity, periodicity and random volatility. The traditional grey model has been difficult to appropriately predict its future evolution. This paper focuses on the adaptability of the model, optimizes and improves the accumulation operator and model structure, and establishes a fractional-order structural self-adaptation grey Bernoulli model based on new information priority. Firstly, based on the new information priority accumulation operator, it is extended to the fractional order. In terms of model structure, combined with NGBM(1,1) model, SADGM(1,1) model and FPGM(1,1) model, periodic fluctuation term and nonlinear power term are included to improve the model's capacity to capture nonlinear, fluctuating and periodic features, and enhance the adaptability and flexibility of the model. The backward difference technique yields the model's parameter estimation and temporal response sequence. Based on the results of the algorithm comparison experiment, the Improved Grey Wolf Optimization Algorithm is chosen to optimize the structural parameters of the model in an effort to enhance its performance. The performance comparison experiment of the model was designed, and three cases of China's hydropower generation, China's renewable energy power generation installed capacity and China's solar energy quarterly power generation were selected to compare the performance with a variety of grey prediction models. Monte-Carlo simulation and probability density analysis were utilized to confirm the stability and accuracy of the proposed model. The results show that the proposed FSANGBM(1,1) model can handle data series of renewable energy with nonlinear, volatile, and periodic features with high prediction ability. Finally, the model is applied to forecast three cases' future development trends.