Численные методы и анализ данных. Рубрика в журнале - Компьютерная оптика
"Экзотические" бинарные системы счисления для колец целых чисел Гаусса и Эйзенштейна
Статья научная
В работе рассматриваются нестандартные бинарные системы счисления для колец целых чисел Гаусса и Эйзенштейна. Принципиальным отличием («экзотичностью») таких систем счисления от канонических систем счисления И. Катаи для квадратичных полей является использование в качестве бинарного «цифрового алфавита» двухэлементного множества, не содержащего числового нуля. В работе синтезируются также алгоритмы представления чисел в рассматриваемой системе счисления и характеризуются возможности эффективной реализации арифметических операций.
Бесплатно
A novel approach for partial shape matching and similarity based on data envelopment analysis
Статья научная
Due to the growing number of 3D objects in digital libraries, the task of searching and browsing models in an extensive 3D database has been the focus of considerable research in the area. In the last decade, several approaches to retrieve 3D models based on shape similarity have been proposed. The majority of the existing methods addresses the problem of similarity between objects as a global matching problem. Consequently, most of these techniques do not support a part of the object as a query, in addition to their poor performance for classes with globally non-similar shape models and also for articulated objects. The partial matching technique seems to be a suitable solution to these problems. In this paper, we address the problem of shape matching and retrieval. We propose a new approach based on partial matching in which each 3D object is segmented into its constituent parts, and shape descriptors are computed from these elements to compare similarities. Several experiments investigated that our technique enables fast computing for content-based 3D shape retrieval and significantly improves the results of our method based on Data Envelopment Analysis descriptor for global matching.
Бесплатно
Analysis of logistics distribution path optimization planning based on traffic network data
Статья
With the development of economy, the distribution problem of logistics becomes more and more complex. Based on the traffic network data, this study analyzed the vehicle routing problem (VRP), designed a dynamic vehicle routing problem with time window (DVRPTW) model, and solved it with genetic algorithm (GA). In order to improve the performance of the algorithm, the genetic operation was improved, and the output solution was further optimized by hill climbing algorithm. The analysis of example showed that the improved GA algorithm had better performance in path optimization planning, the total cost of planning results was 31.44 % less than that of GA algorithm, and the total cost of planning results increased by 11.48 % considering the traffic network data. The experimental results show that the improved GA algorithm has good performance and can significantly reduce the cost of distribution and that research on VRP based on the traffic network data is more in line with the actual situation of logistics distribution, which is conducive to the further application of the improved GA algorithm in VRP.
Бесплатно
Статья научная
With the development of computer technology, there are more and more algorithms and models for data processing and analysis, which brings a new direction to radar target recognition. This study mainly analyzed the recognition of high resolution range profile (HRRP) in radar target recognition and applied the generalized regression neural network (GRNN) model for HRRP recognition. In order to improve the performance of HRRP, the fruit fly optimization algorithm (FOA) algorithm was improved to optimize the parameters of the GRNN model. Simulation experiments were carried out on three types of aircraft. The improved FOA-GRNN (IFOA-GRNN) model was compared with the radial basis function (RBF) and GRNN models. The results showed that the IFOA-GRNN model had a better convergence accuracy, the highest average recognition rate (96.4 %), the shortest average calculation time (275 s), and a good recognition rate under noise interference. The experimental results show that the IFOA-GRNN model has a good performance in radar target recognition and can be further promoted and applied in practice.
Бесплатно
Arrhythmia detection using resampling and deep learning methods on unbalanced data
Статья научная
Due to cardiovascular diseases millions of people die around the world. One way to detect abnormality in the heart condition is with the help of electrocardiogram signal (ECG) analysis. This paper’s goal is to use machine learning and deep learning methods such as Support Vector Machines (SVM), Random Forests, Light Gradient Boosting Machine (LightGBM), Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM) and Bidirectional Long Short-Term Memory (BLSTM) to classify arrhythmias, where particular interest represent the rare cases of disease. In order to deal with the problem of imbalance in the dataset we used resampling methods such as SMOTE Tomek-Links and SMOTE ENN to improve the representation ration of the minority classes. Although the machine learning models did not improve a lot when trained on the resampled dataset, the deep learning models showed more impressive results. In particular, LSTM model fitted on dataset resampled using SMOTE ENN method provides the most optimal precision-recall trade-off for the minority classes Supraventricular beat and Fusion of ventricular and normal beat, with recall of 83 % and 88 % and precision of 74 % and 66 % for the two classes respectively, whereas the macro-weighted recall is 92 % and precision is 82 %.
Бесплатно
Convolutional Neural Network-Based Low Light Image Enhancement Method
Статья научная
With advances in science and technology, remote sensing images are vital for vegetation monitoring. The use of remote sensing allows for the collection of widespread, multi-temporal data on vegetation, leading to a better comprehension and management of natural resources. In this study, a new remote sensing image recognition model is proposed by combining the filtering algorithm to reconstruct the time series curve, fusing the quadratic difference method and decision tree, and introducing the morphological similarity distance method. The results of the experiment indicate that the normalized vegetation index in towns was consistently higher than the index in bodies of water throughout the year. The normalized index for water was generally close to or below 0. Additionally, the normalized index for forests surpassed that of both water and towns. Although the waveforms for all three were similar, the differences were significant. Notably, the forest normalized index curves had a single peak with a noteworthy duration. The study found that the mapping accuracy for fall plants was highest in 2013 (97.37 %) and lowest in 2014 (80.00 %). Similarly, for spring vegetation, the mapping accuracy was greatest in 2017 (97.96 %) and lowest in 2014, but still favorable at 90.00 %. These results highlight the high advantages of the remote sensing identification method proposed by the study for vegetation identification, which is crucial for natural resource management and environmental protection.
Бесплатно
Data mining of corporate financial fraud based on neural network model
Статья научная
Under the active market economy, more and more listed companies emerge. Because of the various interest relationships faced by listed companies, some enterprises which are not well managed or want to enhance company’s value will choose to forge financial reports by improper means. In order to find out the false financial reports as accurately as possible, this paper briefly introduced the relevant indicators for judging the fraudulence of financial reports of listed companies and the recognition model of financial reports based on back propagation (BP) neural network. Then the selection of the input relevant indexes was improved. The improved BP neural network was simulated and analyzed in MATLAB software and compared with the traditional BP neural network and support vector machine (SVM). The results showed that the importance of total assets net profit, earnings per share, cash reinvestment rate, operating gross profit and pre-tax ratio of profit to debt was the top 5 among 20 judgment indexes. In the identification of testing samples of financial report, the accuracy, precision, recall rate and F value all showed that the performance of the improved BP neural network was better than that of the traditional BP network and SVM.
Бесплатно
Fine-tuning the hyperparameters of pre-trained models for solving multiclass classification problems
Статья научная
This study is devoted to the application of fine-tuning methods for Transfer Learning models to solve the multiclass image classification problem using the medical X-ray images. To achieve this goal, the structural features of such pre-trained models as VGG-19, ResNet-50, InceptionV3 were studied. For these models, the following fine-tuning methods were used: unfreezing the last convolutional layer and updating its weights, selecting the learning rate and optimizer. As a dataset chest X-Ray images of the Society for Imaging Informatics in Medicine (SIIM), as the leading healthcare organization in its field, in partnership with the Foundation for the Promotion of Health and Biomedical Research of Valencia Region (FISABIO), the Valencian Region Medical ImageBank (BIMCV) ) and the Radiological Society of North America (RSNA) were used. Thus, the results of the experiments carried out illustrated that the pre-trained models with their subsequent tuning are excellent for solving the problem of multiclass classification in the field of medical image processing. It should be noted that ResNet-50 based model showed the best result with 82.74 % accuracy. Results obtained for all models are reflected in the corresponding tables.
Бесплатно
High-speed recursive-separable image processing filters
Статья научная
The development of modern technologies in the field of image formation leads to an increase in the size of the generated images, as a result the question of reducing the processing computational costs arises, and this is an important factor in the creation of real-time systems. The study provides a description of high-speed recursive-separable filters for improving the quality of images, which, due to the peculiarities of their implementation, can reduce the number of computational operations required for the image processing process. This type of filters is obtained from two-dimensional linear digital filters, which are modified by applying recursive and separable properties to them. The MATLAB environment computing method for implementation of these filters is described. An extensive performance research of the developed filters has been carried out at various sizes of the test image and on various experimental installations. The comparison with the classical two-dimensional convolution method of the developed filters is demonstrated, and it shows the time gain required for the image processing. The results obtained can be applied in biomedical image processing systems or in vision systems working in heavy weather conditions.
Бесплатно
Human Action Recognition Based on The Skeletal Pairwise Dissimilarity
Статья научная
The main idea of the paper is to apply the principles of featureless pattern recognition to human activity recognition problem. The article presents the human figure representing approach based on pairwise dissimilarity function of skeletal models and a set of reference objects, also known as a basic assembly. The paper includes a basic assembly analysis and we propose the method for selecting the least-correlated basic objects. The video sequence proposed for analysis of human activity within frames is represented as an activity map. The activity map is a result of computing the pairwise dissimilarity function between skeletal models from the video sequence and the basic assembly of skeletons. The paper conducts frame-by-frame annotation of activities in the TST Fall Detection v2 database, such as standing, sitting, lying, walking, falling, post-fall lying, grasp, ungrasp. A convolutional neural network based on the ResNetV2 with the SE-block is proposed to solve the activity recognition problem. SE-block allows to detect inter-channel dependencies and selecting the most important features. Additionally, we prepare a data for training, determine an optimal hyperparameters of the neural network model. Experimental results of human activity recognition on the TST Fall Detection v2 database using the Leave-one-person-out procedure are provided. Furthermore, the paper presents a frame-by-frame assessment of the quality of human activity recognition, achieving an accuracy exceeding 83%.
Бесплатно
Illustration visual communication based on computer vision image retrieval algorithm
Статья научная
In illustration design, good visual communication can make the audience resonate. Computer vision image retrieval algorithm provides important support and assistance for the visual communication of illustration. However, the traditional image retrieval algorithm has problems of subjectivity and inaccuracy in complex image classification. Therefore, this paper optimizes the feature extraction module of convolutional neural network and fuses hash algorithm to improve the efficiency and speed of image retrieval. The experimental results show that the accuracy of the improved convolutional neural network is 82.7 %, which is more than 6 percentage points higher than the traditional algorithm model. The recall rate of the volume neural network model improved by hashing algorithm is 94.1 %. Research is of great significance to the visual communication of illustration design, which helps designers to find relevant materials more accurately, improve the artistic quality and ornamental value of their works, and promote the innovation and development of illustration design.
Бесплатно
Статья научная
Forest characteristics estimation is a vital task for ecological monitoring and forest management. Forest owners make decisions based on timber type and its quality. It usually requires field based observations and measurements that is time- and labor-intensive especially in remote and vast areas. Remote sensing technologies aim at solving the challenge of large area monitoring by rapid data acquisition. To automate the data analysis process, machine learning (ML) algorithms are widely applied, particularly in forestry tasks. As ground truth values for ML models training, forest inventory data are usually leveraged. Commonly it involves individual forest stand measurements that are less precise than sample plots. In this study, we delve into ML-based solution development to create spatial-distributed maps with volume stock using sample plot measurements as reference data. The proposed pipeline includes medium-resolution freely available Sentinel-2 data. The experiments are conducted in the Perm region, Russia, and show a high capacity of ML application for forest volume stock estimation based on multispectral satellite observations. Gradient boosting achieves the highest quality with MAPE equal to 30.5%. In future, the proposed solution can be used by forest owners and integrated in advanced systems for ecological monitoring.
Бесплатно
Статья научная
This paper explores the integration of Residual Networks (ResNets) into the in-loop filtering (ILF) process of the Versatile Video Coding (VVC) standard, aiming to enhance video compression efficiency and video quality through the application of Deep Convolutional Neural Networks (DCNNs). The study introduces a novel architecture, the Residual Deep Convolutional Neural Network (RDCNN), designed to replace conventional VVC in-loop filtering modules, including Deblocking Filter (DBF), Sample Adaptive Offset (SAO), and Adaptive Loop Filter (ALF). By leveraging the Rate Distortion Optimization (RDO) technique, the RDCNN model is applied to every coding unit (CU) to optimize the balance between video quality and bitrate. The proposed methodology involves offline training with specific parameters using the TensorFlow-GPU platform, followed by feature extraction and prediction of optimal filtering decisions for each video frame during the encoding process. The results demonstrate the effectiveness of the proposed RDCNN in significantly reducing the bitrate while maintaining high visual quality, outperforming existing methods in terms of compression efficiency and peak signal-to-noise ratio (PSNR) values across various video files (YUV color space). Specifically, the RDCNN achieved a YUV PSNR of 41.2 dB and a BD-rate reduction of – 2.43% for the Y component, – 6.96% for the U component, and – 9.43% for the V component. These results underscore the potential of deep learning techniques, particularly ResNets, in addressing the complexities of video compression and enhancing the VVC standard. The evaluation across various YUV video files, including Stefan_cif, Soccer, Mobile, Harbour, Crew, and Bus, revealed consistently higher average YUV PSNR values compared to both VTM 22.2 and other related methods. This indicates not only improved compression efficiency but also enhanced visual quality, crucial for diverse video processing tasks.
Бесплатно
Статья научная
An automatic speech recognition system has the possibility of enhancing the standard of living for persons with disabilities by solving issues such as dysarthria, stuttering, and other speech defects. In this paper, we introduce a voice assistant using hyperkinetic dysarthria (HD) defect speeches. It contains the data preprocessing steps and the development of a novel convolutional recurrent network (CRN) model that is built depending on the convolutional neural networks and recurrent neural networks. We implemented data preprocessing methods, including filtering, down-sampling, and splitting, to prevent overfitting and decrease processing power as well as time. In addition, the technique of Mel Frequency Cepstral Coefficients (MFCC) has been utilized to extract speech characteristics. The proposed model is trained to recognize HD speech disorders using a dataset including 2000 Russian speeches. The experimental results demonstrate that the proposed method obtains a character error rate (CER) of 14.76 %. It indicates that approximately 85 % of characters are able to correctly recognize on the test dataset. We have created a telegram bot that utilizes our trained model to help people with hyperkinetic dysarthria speech disorder. This bot is capable of providing assistance independently, without the need for any third-party assistance.
Бесплатно
Many heads but one brain: fusionbrain - a single multimodal multitask architecture and a competition
Статья научная
Supporting the current trend in the AI community, we present the AI Journey 2021 Challenge called FusionBrain, the first competition which is targeted to make a universal architecture which could process different modalities (in this case, images, texts, and code) and solve multiple tasks for vision and language. The FusionBrain Challenge combines the following specific tasks: Code2code Translation, Handwritten Text recognition, Zero-shot Object Detection, and Visual Question Answering. We have created datasets for each task to test the participants' submissions on it. Moreover, we have collected and made publicly available a new handwritten dataset in both English and Russian, which consists of 94,128 pairs of images and texts. We also propose a multimodal and multitask architecture - a baseline solution, in the centre of which is a frozen foundation model and which has been trained in Fusion mode along with Single-task mode. The proposed Fusion approach proves to be competitive and more energy-efficient compared to the task-specific one.
Бесплатно
Many-parameter m-complementary Golay sequences and transforms
Статья научная
In this paper, we develop the family of Golay–Rudin–Shapiro (GRS) m-complementary many-parameter sequences and many-parameter Golay transforms. The approach is based on a new gen-eralized iteration generating construction, associated with n unitary many-parameter transforms and n arbitrary groups of given fixed order. We are going to use multi-parameter Golay transform in Intelligent-OFDM-TCS instead of discrete Fourier transform in order to find out optimal values of parameters optimized PARP, BER, SER, anti-eavesdropping and anti-jamming effects.
Бесплатно
Статья научная
An original approach to solving difficult time-consuming problems of registration and analysis of random point images is described. The approach is based on the development and application of high-performance specialized computer algebra systems. Three software packages have been created specifically for carrying out equivalent analytical transformations on a computer. The first software system is designed to calculate formulas describing the volumes of convex polyhedra with parametrically specified boundaries in n -dimensional space. The second system is based on the calculation of multidimensional integral expressions by the method of cyclic differentiation of the integral with respect to the parameter. The third system is based on the accelerated implementation of complex combinatorial-recursive transformations on a computer. Another distinctive feature of the work is the extension of the classical Catalan numbers to the multidimensional case (they were required to solve a number of intermediate probabilistic-combinatorial problems). The implementation of the above computer algebra software systems on a multi-core cluster of Novosibirsk State University, together with the direct use of the explicit form of generalized Catalan numbers, allowed the authors to obtain several new previously unknown probabilistic formulas and dependencies required for solving problems in the field of analysis of random point images.
Бесплатно
Статья научная
Beam divergence is one of the instrument resolution parameters in neutron computed tomography. In pinhole geometry, due to the finite size of the source, geometric unsharpness affects the transmission images and therefore influences the reconstructed data. In this paper, we propose an approach for deterministic simulation of this effect for a voxelized 3D object. The idea behind the proposed approach is to use multiple point sources at a pinhole position and collect transmission images from each of them. The implementation was done using the ASTRA toolbox by calculating cone beam projections from each point source. This approach was applied to a porous phantom. Artifacts associated with beam divergence were identified in the reconstructed data. The influence of beam divergence on the segmentation of pores by binarization of the reconstructed data has been considered.
Бесплатно
Multigrammatical modelling of neural networks
Статья научная
This paper is dedicated to the proposed techniques of modelling artificial neural networks (NNs) by application of the multigrammatical framework. Multigrammatical representations of feed-forward and recurrent NNs are described. Application of multiset metagrammars to modelling deep learning of NNs of the aforementioned classes is considered. Possible developments of the announced approach are discussed.
Бесплатно
Network community partition based on intelligent clustering algorithm
Статья научная
The division of network community is an important part of network research. Based on the clustering algorithm, this study analyzed the partition method of network community. Firstly, the classic Louvain clustering algorithm was introduced, and then it was improved based on the node similarity to get better partition results. Finally, experiments were carried out on the random network and the real network. The results showed that the improved clustering algorithm was faster than GN and KL algorithms, the community had larger modularity, and the purity was closer to 1. The experimental results show the effectiveness of the proposed method and make some contributions to the reliable community division.
Бесплатно
