International Journal of Pattern Recognition and Artificial Intelligence

When treating a brain tumor, a doctor needs to know the site and the size of the tumor. Positron emission tomography (PET) can be effectively applied to diagnose such cancers based on the heightened glucose metabolism of early-stage cancer cells. The purpose of this research is to extract the regions of skull, brain tumor, and brain tissue from a series of PET brain images and then a three-dimensional (3D) model is reconstructed from the extracted skulls, brain tumors, and brain tissues. Knowing the relative site and size of a tumor within the skull is helpful to a doctor. The contours obtained by the segmentation method proposed in this study are quantitatively compared with the contours drawn by doctors on the same image set since the ground truth is unknown. The experimental results are impressive.

Being a class of effective feature descriptors for action recognition, action representations based on skeleton sequences have yielded excellent recognition results. Most methods used to construct these action representations are based on the information from all the joint positions in actions. Unfortunately, some joints in the actions do not improve the accuracy of action recognition, and may even cause unnecessary inter-class errors. In this study, the authors propose a new method for action recognition by selecting active joints which are closely related to the movement of the body as the first step. Further, a skeleton is characterized as a set of its active-joint positions, and the set can be mapped to a point on pre-shape space to filter out the scale and translation variability. Then, a skeleton sequence (an action) can be regarded as points on the space. Because the timing-sequence relationship between skeletons is very valuable for action recognition, a tensor-based linear dynamical system (tLDS) is employed to model the temporal information of the action. To avoid using a finite-order sequence to estimate the infinite-order feature descriptor of a tLDS, the descriptor is mapped to a point on an infinite Grassmannian composed of the extended observability subspaces. The action is classified using sparse coding and dictionary learning (SCDL) on the infinite Grassmannian. Experimental results demonstrate that the recognition accuracies of the proposed method outperform state-of-the-art ones on four different action datasets.

Traffic prediction is a classical time series prediction which has been investigated in different domains, but most existing models are proposed based on limited time or spatial scale. Mobile cellular network traffic prediction is of paramount importance for quality-of-service (QoS) and power management of the cellular base stations, especially in the 5G era. Through the statistical analysis of the real historical traffic data obtained in a city scale spanning across multiple months, this paper makes an in-depth study of the temporal characteristics and behavior rules of the model data traffic. Considering that the time series data show different changing rules under the different time dimensions, spatial dimensions and independent dimensions, a multi-dimensional recurrent neural network (MDRNN) prediction model is established to predict the future cell traffic volume over various temporal and spatial dimensions. The data of this paper are trained and tested over real data of a city, and the granularity of the proposed prediction model can be drilled down to the cell level. Compared with the traditional trend fitting method, the proposed model achieves mean absolute percentage error (MAPE) reduction of 6.56%, and provides guidance for energy efficiency optimization and power consumption reduction of base stations in various temporal and spatial dimensions.

Graph pattern matching that aims to seek out answer graphs in a data graph matching a provided graph, plays a fundamental role as a part of graph search for graph databases. “Matching” indicates that the two graphs are correlated, such as bisimulation, isomorphism, simulation, etc. The strictness of bisimulation is between simulation and isomorphism. Seldom work has been done to search for bisimulation subgraphs. This research focuses on the problem. The symbol $\ast$ is introduced to fundamental modal logic language, thereby yielding $ML+\ast$ language; the symbols ${\square }^{\ast }$ is added for forming $ML+\ast$ formulas. Then conclusions about graph bisimulations are shown. Subsequently, a theorem with detailed proof is presented, stating that $ML+\ast$ formulas characterize finite directed graphs modulo bisimulation. According to the conclusions and theorem, algorithms for finding subgraphs are proposed. After dividing the query graph, the match graphs undergo the characterization using $ML+\ast$ formulas. In the data graphs, by model checking the formulas, the answer graphs exhibiting bisimilarity to the match graphs are able to be captured.

There are limited coronavirus disease 2019 (COVID-19) testing kits, therefore, development of other diagnosis approaches is desirable. The doctors generally utilize chest X-rays and Computed Tomography (CT) scans to diagnose pneumonia, lung inflammation, abscesses, and/or enlarged lymph nodes. Since COVID-19 attacks the epithelial cells that line our respiratory tract, therefore, X-ray images are utilized in this paper, to classify the patients with infected (COVID-19 $+$ve) and uninfected (COVID-19 $-$ve) lungs. Almost all hospitals have X-ray imaging machines, therefore, the chest X-ray images can be used to test for COVID-19 without utilizing any kind of dedicated test kits. However, the chest X-ray-based COVID-19 classification requires a radiology expert and significant time, which is precious when COVID-19 infection is increasing at a rapid rate. Therefore, the development of an automated analysis approach is desirable to save the medical professionals’ valuable time. In this paper, a deep convolutional neural network (CNN) approach is designed and implemented. Besides, the hyper-parameters of CNN are tuned using Multi-objective Adaptive Differential Evolution (MADE). Extensive experiments are performed by considering the benchmark COVID-19 dataset. Comparative analysis reveals that the proposed technique outperforms the competitive machine learning models in terms of various performance metrics.

Particle swarm optimization (PSO) algorithm is a stochastic and population-based optimization algorithm. Its traditional learning strategy is implemented by updating the best position using the particle’s own historical best experience and its neighborhood’s best experience to find the optimal solution of the problem. However, the learning strategy is ineffective when dealing with highly complex problems. In this paper, a particle swarm optimization algorithm based on a multidimensional mean learning strategy is proposed. In this algorithm, an opposition-based learning strategy is utilized to initialize the population to enhance the exploitation capability. Furthermore, the historical best positions of all the particles are reconstructed in a vertical crossover manner that is based on the mean information of multiple optimal dimensions to generate the guiding particles. Additionally, an improved inertia weight is used to further guide all the particle movements to balance the capability of the proposed algorithm for global exploration and local exploitation. The proposed algorithm is tested on 12 benchmark functions and is compared with some well-known PSO algorithms. The experimental results show that the proposed algorithm obtains more competitive optimal solution compared with other PSO algorithms when solving high-dimensional complex problems.

Single-image super-resolution (SR) imaging is a fundamental problem in image processing; it is important in entertainment, video surveillance, remote sensing, medicine, and other fields. Gaussian process regression (GPR) is a kernel method whereby nonlinear mapping relationships in data can be learned. However, the traditional Gaussian kernel function used in GPR is isotropic and fails to capture complex image structures. Accordingly, the structure information of image patches, termed steering kernel coefficients (SKCs), is extracted by a steering kernel function. After patches with similar structure are clustered according to their SKCs, an anisotropic automatic-relevance-determination (ARD) kernel function is used to learn the model for each cluster. Aiming at learning a structure-sensitive GPR model, we integrate the SKCs and ARD to achieve improved performance for GPR-based SR. Experiments demonstrate that the proposed method can effectively capture the structural relevance of image patches and yield promising results.

Point cloud segmentation is a crucial fundamental step in 3D reconstruction, object recognition and scene understanding. This paper proposes a supervoxel-based point cloud segmentation algorithm in region growing principle to solve the issues of inaccurate boundaries and nonsmooth segments in the existing methods. To begin with, the input point cloud is voxelized and then pre-segmented into sparse supervoxels by flow constrained clustering, considering the spatial distance and local geometry between voxels. Afterwards, plane fitting is applied to the over-segmented supervoxels and seeds for region growing are selected with respect to the fitting residuals. Starting from pruned seed patches, adjacent supervoxels are merged in region growing style to form the final segments, according to the normalized similarity measure that integrates the smoothness and shape constraints of supervoxels. We determine the values of parameters via experimental tests, and the final results show that, by voxelizing and pre-segmenting the point clouds, the proposed algorithm is robust to noises and can obtain smooth segmentation regions with accurate boundaries in high efficiency.

Underwater image capturing is a challenging task due to attenuation of light in water. Scattering and absorption are the results of light attenuation which lead to faded colors and reduced contrast of images, respectively. To deal with these issues and to provide better visual quality image, various enhancement methods have been proposed. This paper proposes the Dual Domain-based Underwater Image Enhancement (DDUIE) method. DDUIE method provides contrast stretching in approximation band of discrete wavelet transformed image followed by intensity adjustment of different color channels in spatial domain. To further improve the color quality, the image is processed in HSV (Hue–Saturation–Value) color space. Result analysis indicates better results for DDUIE method over state-of-the-art methods. Subjective results of DDUIE method show minimization of the bluish-green effect and reduction of nonuniform illumination up to a certain extent. These lead to enhanced color and image details. Quantitative results show that the Underwater Image Quality Measure (UIQM) and Underwater Color Image Quality Evaluation (UCIQE) values between 1 and 2 and between 0 and 1 have been achieved, respectively, which significantly illustrate that images have been enhanced efficiently and also entropy values between 7 and 8 depict the effectiveness of the proposed method in terms of image details.

Aiming at the problems of low segmentation accuracy of noise image, poor noise immunity of the existing models and poor adaptability to complex noise environment, a noise image segmentation algorithm using anisotropic diffusion and nonconvex functional was proposed. First, focusing on the “staircase effect”, a nonconvex functional was introduced into the energy functional model for smooth denoising. Second, the validity and accuracy of the model were established by proving that there was no global minimum in the solution space of the nonconvex energy functional model; the improved model was then used to obtain a smooth and clear image edge while maintaining the edge integrity. Third, the smooth image obtained from the nonconvex energy functional model was combined with the level set model to obtain the anisotropic diffusion gray level set model. The optimal outline of the target was obtained by calculating the minimum value of the energy functional. Finally, an anisotropic diffusion equation with nonconvex energy functional model was built in this algorithm to segment noise image accurately and quickly. A series of comparative experiments on the proposed algorithm and similar algorithms were conducted. The results showed that the proposed algorithm had strong noise resistance and provided precise segmentation for noise image.

Innovative applications in rapidly evolving domains such as robotic navigation and autonomous (driverless) vehicles rely on binocular computer vision systems that meet stringent response time and accuracy requirements. A key problem in these vision systems is stereo matching, which involves matching pixels from two input images in order to construct the output, a 3D map. Building upon the existing local stereo matching algorithms, this paper proposes a novel stereo matching algorithm that is based on a weighted guided filtering foundation. The proposed algorithm consists of three main steps; each step is designed with the goal of improving accuracy. First, the matching costs are computed using a unique combination of complementary methods (absolute difference, Census, and gradient algorithms) to reduce errors. Second, the costs are aggregated using an adaptive weighted guided image filtering method. Here, the regularization parameters are adjusted adaptively using the Canny method, further reducing errors. Third, a disparity map is generated using the winner-take-all strategy; this map is subsequently refined using a densification method to reduce errors. Our experimental results indicate that the proposed algorithm provides a higher level of accuracy in comparison to a collection of the existing state-of-the-art local algorithms.

An important work of data analysis is to identify correlation structures and classify the data in unlabeled high-dimensional data, which usually requires iterative experiments on clustering parameters, attribute weights and instances. For a large dataset, the number of clusters may be huge, and it is a great challenge to explore in this huge space. People usually have a more comprehensive understanding of some data. For example, they think that data A is better than data B, but they do not know which attributes are important. Therefore, a powerful interactive analysis tool can help people greatly improve the effectiveness of exploratory clustering analysis. This paper provides a visual analysis method for sorting and classifying multivariate data. It can determine the weight of each attribute through user’s interaction, thus, generating sorting, and then complete classification according to sorting results. Through visual display, users can understand the characteristics of data as well as category characteristics intuitively and quickly, and it helps users improve sorting and classification results.

The aim of this paper is to objectively present the development and trend of international MOOC research. Metadata taken from 660 literature records is used to visualize the state of MOOC research. The records were published between 2008 and 2018, and Citespace software is employed to visualize key data about the research contained on the SCI and SSCI platform. It has the following findings. First, since 2012, the number of international MOOC research papers and cited frequency has shown an upward trend. The research force was mainly concentrated in North America, Europe and Asia, and the publishing institutions were mainly concentrated in Universities of various countries, and formed a more obvious international cooperation network. Second, journals such as $\ll$INT REV RES OPEN DIS$\gg$ and $\ll$COMPUTERS & EDUCATION$\gg$ are the foundational literatures of MOOC research. George Siemens, Stephen Downes, John Daniel and others have made important contributions to the foundation research of MOOC. Third, Breslow, Liyanagunawardena et al. were cited more frequently and formed eight research clusters such as teaching interaction, relativism/behaviorism theory, individualized learning and diversified education. Fourth, distance education, online learning, learning situation analysis, curriculum construction and platform construction are the hot topics of international MOOC research in the recent decade. Fifth, the development of international MOOC has spread to various disciplines and promoted the interdisciplinary research of environment, biology, medicine, philosophy, psychology and other disciplines.

Satisfiability problem is a famous nondeterministic polynomial-time complete (NP-complete) problem, which has always been a hotspot in artificial intelligence. In this paper, by combining the advantages of DNA origami with hybridization chain reaction, a computing model was proposed to solve the satisfiability problem. For each clause in the given formula, a DNA origami device was devised. The device corresponding to the clause was capable of searching for assignments that satisfied the clause. When all devices completed the search in parallel, the intersection of these satisfying assignments found must satisfy all the clauses. Therefore, whether the given formula is satisfiable or not was decided. The simulation results demonstrated that the proposed computing model was feasible. Our work showed the capability of DNA origami in architecting automatic computing device. The paper proposed a novel method for designing functional nanoscale devices based on DNA origami.

Traditional template matching-based motion estimation is a popular but time-consuming method for vibration vision measurement. In this study, the particle swarm optimization (PSO) algorithm is improved to solve this time-consumption problem. The convergence speed of the algorithm is increased using the adjacent frames search method in the particle swarm initialization process. A flag array is created to avoid repeated calculation in the termination strategy. The subpixel positioning accuracy is ensured by applying the surface fitting method. The robustness of the algorithm is ensured by applying the zero-mean normalized cross correlation. Simulation results demonstrate that the average extraction error of the improved PSO algorithm is less than 1%. Compared with the commonly used three-step search algorithm, diamond search algorithm, and local search algorithm, the improved PSO algorithm consumes the least number of search points. Moreover, tests on real-world image sequences show good estimation accuracy at very low computational cost. The improved PSO algorithm proposed in this study is fast, accurate, and robust, and is suitable for plane motion estimation in vision measurement.