Extraction Capability Research Articles

Creating an ERAS Audit System with Automated Data Extraction Capabilities to Reduce Overall Data Abstraction Time.

Creating an ERAS Audit System with Automated Data Extraction Capabilities to Reduce Overall Data Abstraction Time.

Ensemble Techniques in Plant Nutrient Deficiency Detection: A Survey

Abstract: Plant nutrient deficiency detection represents a critical challenge in modern agriculture, significantly im- pacting crop health, yield, and sustainable farming practices. This survey paper presents a comprehensive analysis of ensemble learning techniques for automated detection of plant nutrient deficiencies through image analysis. Our research evaluates a hybrid approach combining MobileNet’s efficient feature extraction capabilities with ensemble methods including Random Forest and Gradient Boosting algorithms. The methodology encompasses three key compo- nents: feature extraction using MobileNet’s lightweight architecture for processing plant leaf images, implementation of individual classification models, and development of ensemble techniques that leverage the strengths of multiple clas- sifiers. Through extensive experimentation and comparative analysis, our findings demonstrate that ensemble methods consistently outperform individual models, achieving superior accuracy in detecting nutrient deficiencies across various plant species. The integration of MobileNet with ensemble techniques provides a robust framework that balances com- putational efficiency with prediction accuracy. This research contributes to the advancement of precision agriculture by proposing a scalable, real-time solution for nutrient deficiency detection that can be practically implemented in field conditions, ultimately supporting improved crop management decisions and agricultural productivity..

Fault diagnosis for ion mill etching machine cooling system based on omni-dimensional dynamic convolution and dynamic spatial pyramid pooling

ABSTRACT Ion mill etching machine is a critical equipment in semiconductor manufacturing for wafer etching processes. The cooling system, as an integral part of the Ion mill etching machine, is prone to malfunctions during operation, which can result in the scrapping of subsequent wafers processed by the equipment. Therefore, early detection of cooling system faults is crucial to prevent wafer wastage. Due to the high flexibility of the processing process in ion beam etching machines, the length of time-series signals collected by sensors varies significantly, which poses a challenge for accurate fault diagnosis of the cooling system in ion mill etching machines. To address these issues, we propose a novel end-to-end hybrid model called ODCNN-DSPP. The main contributions are as follows: (1) Enhancing the convolutional neural network (CNN) model structure by introducing an omni-dimensional dynamic convolutional layer to improve feature extraction capabilities; and (2) Utilising a dynamic spatial pyramid pooling method to mitigate the impact of irrelevant features on the CNN. Using publicly available ion mill etching machine datasets for experiments, the results indicate that the method proposed in this paper outperforms current mainstream fault diagnosis methods, achieving an average fault diagnosis accuracy of 96.55% on the test set.

A Comprehensive Survey of Deep Learning and Ensemble Techniques in Glaucoma Detection

Glaucoma stands as a primary contributor to irreversible blindness, necessitating precise and prompt diagnoses for effective management. Recent progress in deep learning, particularly through the use of ensemble methods involving Convolutional Neural Networks (CNNs), has demonstrated considerable potential in automating the detection of glaucoma by analyzing ocular imaging data, such as fundus and Optical Coherence Tomography (OCT) images. This survey provides a thorough overview of the latest ensemble-based approaches developed for glaucoma detection, emphasizing the advantages of integrating various CNN architectures, including ResNet, VGG, and DenseNet, to enhance feature extraction and classification capabilities. The paper explores current trends in transfer learning, multi-modal data integration, and hybrid methodologies that reinforce the performance and adaptability of ensemble methods in clinical environments. Additionally, it addresses challenges like the necessity for high-quality labeled datasets, model interpretability, and generalization across different populations. By exploring into recent studies, the survey aims to identify limitations in existing systems and propose advancements in ensemble- based glaucoma detection, ultimately offering valuable insights into future research path that can narrow the gap between experimental findings and practical clinical applications.

A Survey on Liver Cancer Detection Using Hyperfusion of CNN and SVM in Machine Learning

Since liver cancer ranks among of the most aggressive renditions of the disease, improving patient outcomes requires early identification. We propose an inventive tactic to liver cancer detection by integrating CNN and SVM. CNNs, known for their powerful feature extraction capabilities, are particularly effective in analysing complex medical images. SVMs, on the other hand, are efficient classifiers that can separate data points in high-dimensional spaces with accuracy. By merging the feature extraction strength of CNN with the classification efficiency of SVM, the proposed model aims to enhance liver cancer detection accuracy and robustness. The experimental results reveal that the fused CNN-SVM model significantly surpasses the performance of standalone CNN and SVM models, achieving a high detection accuracy of 95.2%. This hybrid method offers a promising direction for improving the precision of computer-aided diagnosis systems, contributing to more effective and reliable liver cancer detection methods that can assist healthcare professionals in making timely decisions.

Rhodanine Substitution of Asymmetric Nonfullerene Acceptors for High-Performance Organic Solar Cells.

Asymmetric substitution is acknowledged as a straightforward yet potent approach for the optimization of small molecule acceptors (SMAs), thereby enhancing the power conversion efficiency (PCE) of organic solar cells (OSCs). In this work, we have successfully engineered and synthesized a novel asymmetric SMA, designated as Y6-R, which features a rhodanine-terminated inner side-chain. In devices with PM6 as the polymer donor, the asymmetric Y6-R demonstrated an impressive PCE of 18.62% with an open-circuit voltage (Voc) of 0.863 V, a short-circuit current (Jsc) of 27.89 mA cm-2, and a fill factor (FF) of 77.35%, which is much higher than that of the symmetric SMA Y6-based devices (16.84%). The superior performance for PM6:Y6-R devices can be attributed to a combination of factors, including the upshifted LUMO energy levels, more desired exciton dissociation, collection, and extraction capability, as well as reduced recombination and suppressed Eloss. In addition, the Y6-R molecules promote the more desired coaggregation behaviors along with the donor, endowing the stronger and more ordered crystallinity of blend films. Our findings underscore the effectiveness of rhodanine-substitution, as a representative of "A" units, of the asymmetric SMAs in fine-tuning the coaggregation behavior with the donor, thereby enhancing crystallinity and efficiency.

YOLO-VDS: accurate detection of strawberry developmental stages for embedded agricultural robots

Abstract Detecting the various developmental stages of strawberries in their natural environment is crucial for modern agricultural robots. Existing methods focus on fruit detection but ignore stage classification. Moreover, they usually requiring substantial computational resources and are not suitable for small and low-power embedded platforms. To address this problem, we propose YOLO-VDS, a lightweight model based on YOLOv5s and optimized for embedded platforms. We introduce the Inverse Residual Bottleneck with 3 Convolutions (IRBC3) module to enhance the feature extraction capability and reduce the model computation. Then, we enhance the feature extraction and representation capabilities by introducing Efficient Channel Attention (ECA) in the backbone. Experiments on the Strawberry-DS dataset show that YOLO-VDS performs significantly better than YOLOv5s and other similar algorithms. Compared to YOLOv5s, the accuracy is improved by 5.8%, mAP@ 0.5 by 7.7%, and the number of model parameters is reduced by 24.29%. Deployed on a Jetson TX2 NX, YOLO-VDS reaches 19.2 FPS after TensorRT acceleration, demonstrating its suitability for vision-guided harvesting robots and edge computing applications.

A Universal Visual Detection Method for Camellia oleifera Fruit Picking Robot

ABSTRACTIn recent years, the application of robots in the field of fruit picking has steadily increased. Mechanized methods for harvesting oil tea fruits include comb picking, vibratory picking, and gripping picking, among others. Traditional reliance on a single‐picking method is limited by variability in fruit size, shading, and environmental conditions. To develop a universal vision system suitable for picking robots capable of multiple picking methods and achieve intelligent harvesting of oil tea fruits. This paper proposes an enhanced You Only Look Once v7 (YOLOv7)‐based oil tea fruits recognition method specifically designed for subsequent clamp or comb picking. The network's feature extraction capability is enhanced by incorporating an attention mechanism, an optimized small target detection layer, and an improved training loss function, thereby improving its detection of occluded and small target fruits. An innovative Automatic Assignment (AA) method clusters and subclusters the detected oil tea fruits, providing crucial fruit distribution data to optimize the robot's picking strategy. Additionally, for vibration harvesting, this paper introduces a vibration point detection method utilizing the Pyramid Scene Parsing Network (PSPNet) semantic segmentation network combined with connectivity domain analysis to identify vibration points on the trunks and branches of oil tea trees. Experimental results demonstrate that the generalized visual detection method proposed in this study surpasses existing models in identifying oil tea fruit trees, with the enhanced YOLOv7 model achieving mean average precision, recall, and accuracy of 91.7%, 94.0%, and 94.9%, respectively. The AA method achieves clustering and subclustering of oil tea fruits with a processing delay of under 5 ms. For vibration harvesting, PSPNet achieves branch segmentation precision, recall, and intersection ratio of 97.3%, 96.5%, and 94.5%, respectively. The proposed branch vibration point detection method attains a detection accuracy of 93%, effectively pinpointing vibration points on the trunks and branches of oil tea trees. Overall, the proposed visual method can be implemented in robots using various picking techniques to enable automated harvesting.

QuanFormer: A Transformer-Based Precise Peak Detection and Quantification Tool in LC-MS-Based Metabolomics.

In metabolomic analysis based on liquid chromatography coupled with mass spectrometry, detecting and quantifying intricate objects is a massive job. Current peak picking methods still cause high rates of incorrectly picked peaks to influence the reliability and reproducibility of results. To address these challenges, we developed QuanFormer, a deep learning method based on object detection designed to accurately quantify peak signals. Our algorithm combines the feature extraction capabilities of convolutional neural networks (CNNs) with the global computation capability of Transformer architecture. Data training in QuanFormer by using nearly 20,000 annotated regions-of-interest (ROIs) ensures unique prediction via bipartite matching, achieving 96.5% of the average precision value on the test set. Even without retraining, QuanFormer achieves over 90% accuracy in distinguishing true from false peaks. Performance was further analyzed using visualization techniques applied to the encoder and decoder layers. We also demonstrated that QuanFormer could correct retention time shifts for peak alignment and generally surpass the existing methods, including MZmine 3 and PeakDetective, to obtain a larger number of picked peaks and higher accurate quantification. Finally, we also carried out metabolomic analysis in a clinical cohort of breast cancer patients and utilized QuanFormer to detect and quantify the potential biomarkers. QuanFormer is open-source and available at https://github.com/LinShuhaiLAB/QuanFormer.

Machine Learning Based Finger Print Analysis for Gender Detection: A Review

Gender detection using fingerprint biometrics has emerged as a promising area of research due to its non-intrusive nature and potential applications in biometric identification systems. The procedure can involve multiple steps are the size of finger print and their ridge pattern, minutiae point, machine learning and image processing and accuracy and limitations. This review explores the effectiveness of machine learning techniques for gender classification based on fingerprint patterns, emphasizing the role of advanced classification algorithms and feature extraction methods. Machine learning is crucial for gender detection since it classifies fingerprint patterns and biometric information using models like Convolutional Neural Networks (CNN) and Support Vector Machines (SVM). To identify traits unique to a gender, such as ridge density and minutiae points, these algorithms are trained using labelled datasets. Compared to manual procedures, these models are more effective at handling high-dimensional data and identifying subtle gender-related patterns. Although hybrid models like CNN-DNN and AlexNet further increase classification precision, Convolutional Neural Networks (CNNs) are especially effective due to their automatic feature extraction capabilities. Despite their effectiveness, factors like as picture resolution, demographic balance, and dataset heterogeneity might affect performance, highlighting the need for carefully selected datasets and improved model designs. A structured comparative analysis of multiple studies reveals the impact of various datasets, feature types, and model architectures on classification accuracy and reliability. The findings suggest that deep learning models often outperform traditional classifiers, while dimensionality reduction and hybrid approaches can further enhance performance. However, challenges such as dataset imbalances, limited diversity, and susceptibility to low-quality fingerprint data remain prominent barriers to achieving consistent results. This review also outlines key limitations observed across the studies and provides recommendations for future research, including the need for more diverse datasets and optimized classification frameworks. This study aims to improve fingerprint feature extraction for gender detection, reduce processing costs, fix dataset imbalances, and increase classification accuracy. By stating the objective, the scope and objectives of each investigation are made clear. The generalizability of machine learning models is significantly impacted by the amount, variety, and quality of the dataset. The analysis aims to support the development of more accurate, inclusive, and scalable fingerprint-based gender detection systems.

Enhanced ResNet-50 for garbage classification: Feature fusion and depth-separable convolutions.

As people's material living standards continue to improve, the types and quantities of household garbage they generate rapidly increase. Therefore, it is urgent to develop a reasonable and effective method for garbage classification. This is important for resource recycling and environmental improvement and contributes to the sustainable development of production and the economy. However, existing deep learning-based garbage image classification models generally suffer from low classification accuracy, insufficient robustness, and slow detection speed due to the large number of model parameters. To this end, a new garbage image classification model is proposed, with the ResNet-50 network as the core architecture. Specifically, first, a redundancy-weighted feature fusion module is proposed, enabling the model to fully leverage valuable feature information, thereby improving its performance. At the same time, the module filters out redundant information from multi-scale features, reducing the number of model parameters. Second, the standard 3×3 convolutions in ResNet-50 are replaced with depth-separable convolutions, significantly improving the model's computational efficiency while preserving the feature extraction capability of the original convolutional structure. Finally, to address the issue of class imbalance, a weighting factor is added to the Focal Loss, aiming to mitigate the negative impact of class imbalance on model performance and enhance the model's robustness. Experimental results on the TrashNet dataset show that the proposed model effectively reduces the number of parameters, improves detection speed, and achieves an accuracy of 94.13%, surpassing the vast majority of existing deep learning-based waste image classification models, demonstrating its solid practical value.

Metasurface-based single-pixel recognition through scattering media

Taking advantage of optoelectronic hybrid neural networks, we propose a metasurface-single-pixel hybrid neural network for object recognition. It employs only eight illumination patterns trained by the digital neural network to convolve the object from two-dimensional images into only eight intensity values measured by a single-pixel detector, achieving a 93.8% accuracy rate in handwritten digit recognition. Our work therefore paves an image-free way for metasurface-based object recognition using only a single-pixel detector, which exhibits its powerful information compression and accurate extraction capabilities coupled with a compact structural design.

MammoViT: A Custom Vision Transformer Architecture for Accurate BIRADS Classification in Mammogram Analysis

Background: Breast cancer screening through mammography interpretation is crucial for early detection and improved patient outcomes. However, the manual classification of mammograms using the BIRADS (Breast Imaging-Reporting and Data System) remains challenging due to subtle imaging features, inter-reader variability, and increasing radiologist workload. Traditional computer-aided detection systems often struggle with complex feature extraction and contextual understanding of mammographic abnormalities. To address these limitations, this study proposes MammoViT, a novel hybrid deep learning framework that leverages both ResNet50’s hierarchical feature extraction capabilities and Vision Transformer’s ability to capture long-range dependencies in images. Methods: We implemented a multi-stage approach utilizing a pre-trained ResNet50 model for initial feature extraction from mammogram images. To address the significant class imbalance in our four-class BIRADS dataset, we applied SMOTE (Synthetic Minority Over-sampling Technique) to generate synthetic samples for minority classes. The extracted feature arrays were transformed into non-overlapping patches with positional encodings for Vision Transformer processing. The Vision Transformer employs multi-head self-attention mechanisms to capture both local and global relationships between image patches, with each attention head learning different aspects of spatial dependencies. The model was optimized using Keras Tuner and trained using 5-fold cross-validation with early stopping to prevent overfitting. Results: MammoViT achieved 97.4% accuracy in classifying mammogram images across different BIRADS categories. The model’s effectiveness was validated through comprehensive evaluation metrics, including a classification report, confusion matrix, probability distribution, and comparison with existing studies. Conclusions: MammoViT effectively combines ResNet50 and Vision Transformer architectures while addressing the challenge of imbalanced medical imaging datasets. The high accuracy and robust performance demonstrate its potential as a reliable tool for supporting clinical decision-making in breast cancer screening.

Extraction capability of organic solvents, ionic liquids, and deep eutectic solvents in the extraction of crocin from saffron using ultrasonic assisted extraction optimized by using response surface methodology

Extraction capability of organic solvents, ionic liquids, and deep eutectic solvents in the extraction of crocin from saffron using ultrasonic assisted extraction optimized by using response surface methodology

An Integrated Framework for TFT-LCD Quality Prediction Based on L1 Regularized Feature Selection and SSA-Optimized MLP-XGBoost Ensemble

With the rapid development of display technology, accurate quality prediction of TFT-LCD is crucial because its performance directly affects the user experience. Traditional quality prediction methods rely on manual inspection or rule-based methods, facing the challenges of low efficiency, strong subjectivity, and lack of robustness to changes in production environment. To address these limitations, this paper proposes a novel regression model based on L1 regularized feature selection, Salp Swarm algorithm (SSA) optimization, and MLP-XGBoost integration. L1 regularization is used to automatically select key features from high-dimensional industrial data, reduce redundancy and enhance model generalization. Subsequently, SSA optimizes the hyperparameters of MLP and XGBoost, enabling effective exploration of the parameter space and thus improving model performance. The integrated MLP-XGBoost model combines the nonlinear feature extraction capability of MLP with the regression accuracy of XGBoost. Experimental results show that the proposed model outperforms traditional models in various indicators. Compared with the unoptimized MLP-XGBoost, the model achieves a significant reduction in prediction error, verifying its effectiveness in solving TFT-LCD quality prediction.

MAFF-Net: PCB defect detection via feature-enhancement fusion

Abstract Defect detection of printed circuit board (PCB) is of significant practical importance to ensure quality control in the production process. However, traditional defect detection methods suffer from limitations such as low detection accuracy and poor generalization ability. To tackle these issues, we propose a novel deep learning-based defect detection method for bare PCBs through multi-attention adaptive feature-enhancement fusion (AFF). First, we utilize ResNext101 as the backbone for feature extractor and embed a normalization-based attention mechanism in a residual structure, aiming at improving the feature extraction capability of the network. Second, we introduce an AFF module, which leverages multi-scale feature extraction and feature fusion to facilitate information interaction and enhance the correlation of feature information between channels. Finally, we incorporate the coordinate attention mechanism into AFF to highlight the target area for boosting detection accuracy. The experimental results demonstrate the effectiveness of the proposed method, which achieves a mean accuracy precision (mAP) of 99.01 % on a publicly available PCB defect dataset.

Research on time series analysis of data products in electric power marketing management information platform based on recurrent neural networks

Abstract The power marketing management information platform can promote the energy industry's decision-making ability for power marketing business and improve demand-side management. Due to the large number of power marketing users and heterogeneous data, traditional time series analysis methods are difficult to flexibly deal with this scenario, and deep learning methods based on feedforward neural networks have weak modeling capabilities for the time dimension of data, making it difficult to effectively use these data to support decision-making. To this end, this study proposes a new time series analysis model for the power marketing management information platform, which uses a recurrent neural network (RNN) enhanced by an attention mechanism. By integrating the long-term time dependency modeling capability of RNN and the adaptive extraction capability of the attention mechanism, the model can efficiently represent time series data. Experimental evaluation shows that our method greatly outperforms traditional models and achieves higher prediction accuracy on several key performance indicators.

SED-YOLO based multi-scale attention for small object detection in remote sensing

Object detection is crucial for remote sensing image processing, yet the detection of small objects remains highly challenging due to factors such as image noise and cluttered backgrounds. In response to this challenge, this paper proposes an improved network, named SED-YOLO, based on YOLOv5s. Firstly, we leverage Switchable Atrous Convolution (SAC) to replace the standard convolutions in the original C3 modules of the backbone network, thereby enhancing feature extraction capabilities and adaptability. Additionally, we introduce the Efficient Multi-Scale Attention(EMA) mechanism at the end of the backbone network to enable efficient multi-scale feature learning, which reduces computational costs while preserving crucial information. In the Neck section, an adaptive Concat method is designed to dynamically adjust the feature fusion strategy according to image content and object characteristics, strengthening the model’s ability to handle diverse objects. Lastly, the three-scale feature detection head is expanded to four by adding a small object detection layer, and incorporating the Dynamic Head(DyHead) module. This enhances the detection head’s expressive power by dynamically adjusting attention weights in feature maps. Experimental results demonstrate that this improved network achieves an mean Average Precision (mAP) of 71.6% on the DOTA dataset, surpassing the original YOLOv5s by 2.4%, effectively improving the accuracy of small object detection.

Research on mechanical part recognition method based on improved mask R-CNN instance segmentation

Aiming at the problem of poor part recognition due to mutual occlusion between parts and the influence of different postures in the assembly scene, we propose an improved Mask R-CNN-based part recognition method for complex scenes. Firstly, the ResNet101 network is used to enhance the feature extraction capability of the network and improve the part recognition effect; secondly, the normalization layer of the backbone network is replaced to reduce the effect of batch size on the feature extraction of the model; lastly, the feature pyramid network structure is improved to enhance the transfer efficiency between the high and low layers of the network, and to enhance the capability of the feature capture; through the experiments on the homemade dataset, the average detection accuracy of this method is 4.7 % higher than that of the original Mask R-CNN. Through the experiments on the homemade dataset, it is found that compared with the original Mask R-CNN, the average detection accuracy of the method is improved by 4.7 %.The optimized network model proposed in this paper can improve the accuracy of part recognition, realize the accurate detection of parts in the complex environment such as stacking, occlusion and so on, and provide a solution for the recognition of parts in the complex environment.

DenseNet-Enabled Intelligent Ship Energy Consumption Prediction Method with Navigation Information and Environmental Factors

As a crucial procedure, ship energy consumption prediction is of great significance for the implementation of navigation planning, energy efficiency improvement, and emission reduction strategies. Currently, various studies have been conducted, primarily focusing on container ships, cargo ships, bulk carriers or tankers, with relatively fewer studies dedicated to research vessels. Regarding the limitations of low accuracy and poor effectiveness in ship energy consumption prediction, caused by the diverse voyages and specialized operational missions of research vessels, this paper proposes a DenseNet-enabled energy consumption prediction model considering the complex navigation information and diverse environmental factors. By leveraging the high parameter efficiency of DenseNet, effective feature extraction and strong generalization capabilities can be achieved, which provide more reliable and accurate predictions. Firstly, due to the fact that the real-time fuel consumption is influenced by a variety of internal and external factors, multi-source monitoring data are obtained and analyzed. After data analysis and feature extraction, the processed data and features are utilized to establish the prediction model. Through validations on the actual voyage data, it is demonstrated that the DenseNet-based prediction model outperforms the ResNet, DBCNN, and FCNN models in terms of both accuracy and predictive reliability. Therefore, the proposed method is capable of accurately predicting ship energy consumption under diverse shipping conditions and provides valuable guidance for the development of intelligent shipping.