International Journal of Pattern Recognition and Artificial Intelligence

Curved surface mesh generation is a key step for many areas. Here, a mesh generation algorithm for closed curved surface based on Delaunay refinement is proposed. We focus on improving the shape quality of the meshes generated and making them conform to 2-manifold. The Delaunay tetrahedralization of initial sample is generated first, the initial surface mesh which is a subset of the Delaunay tetrahedralization can be achieved. A triangle is refined by inserting a new point if it is large or of bad quality. For each sample, we also check the triangles that adjoin it whether from a topological disk. If not, the largest triangle will be refined. Finally, the surface mesh is updated after a new point is inserted into the sample. The definition of mesh size function for surface mesh generation is also put in this paper. Meshing experiments of some models demonstrate that the new algorithm is advantageous in generating high quality surface mesh, the count of mesh is suitable and can well approximate the curved surface. The presented method can be used for a wide range of problems including computer graphics, computer vision and finite element method.

Boundary extraction is a fundamental post-clustering problem. It facilitates interpretability and usability of clustering results. Also, it provides visualization and dataset reduction. However, it has not attracted much attention compared to the clustering problem itself. In this work, we address the boundary extraction of clusters in 2- and 3-dimensional spatial datasets. We propose two algorithms based on Delaunay Triangulation (DT). Numerical experiments show that the proposed algorithms generate the cluster boundaries effectively. Also, they yield significant amounts of dataset reduction.

The paper presents a novel approach for classification of faults in Extra High Voltage (EHV) transmission line with series compensation. The proposed algorithm utilizes single end currents extracted from three phases of a transmission line given to a fault detection and classification system. A detailed model is constructed to analyze fault patterns occurring on a dual feed system with multiple series compensation provided on EHV transmission line. The algorithm uses a full cycle of post-fault currents. A Multiresolution Analysis Wavelet-based decomposition technique is used to provide a joint time frequency analysis. Extensive study is done by designing different fault classifiers using Support Vector Machine (SVM) with different feature vector groups. The SVMs with different parameters are compared to show performance of SVMs in given feature space of faults. A new algorithm is proposed with Ensemble method having group of different classifiers, namely, Artificial Neural Network (ANN), K-Nearest Neighborhood (KNN) and SVM. A combination of feature selection with Ensemble Classifiers is trained and tested on a wide range of fault patterns. A significant improvement in performance is obtained with different combination of features in Ensemble Classifier using subspace partition method.The proposed Ensemble Classifier provides an accuracy of 99.5%.

A weighted densely connected convolution network (W-DenseNet) is proposed for reinforcement learning in this work. The W-DenseNet can maximize the information flow between all layers in the network by cross layer connection, which can reduce the phenomenon of gradient vanishing and degradation, and greatly improves the speed of training convergence. The weight coefficient introduced in W-DenseNet, the current layer received all the previous layers’ feature maps with different initial weights, which can extract feature information of different layers more effectively according to tasks. According to the weight adjusted by learning, the cross-layer connection is pruned to remove the cross-layer connection with smaller weight, so as to reduce the number of cross-layer. In this work, GridWorld and FlappyBird games are used for simulation. The simulation results of deep reinforcement learning based on W-DenseNet are compared with the traditional deep reinforcement learning algorithm and reinforcement learning algorithm based on DenseNet. The simulation results show that the proposed W-DenseNet method can make the results more convergent, reduce the training time, and obtain more stable results.

Paraphrase identification is a natural language processing (NLP) problem that involves the determination of whether two text segments have the same meaning. Various NLP applications rely on a solution to this problem, including automatic plagiarism detection, text summarization, machine translation (MT), and question answering. The methods for identifying paraphrases found in the literature fall into two main classes: similarity-based methods and classification methods. This paper presents a critical study and an evaluation of existing methods for paraphrase identification and its application to automatic plagiarism detection. It presents the classes of paraphrase phenomena, the main methods, and the sets of features used by each particular method. All the methods and features used are discussed and enumerated in a table for easy comparison. Their performances on benchmark corpora are also discussed and compared via tables. Automatic plagiarism detection is presented as an application of paraphrase identification. The performances on benchmark corpora of existing plagiarism detection systems able to detect paraphrases are compared and discussed. The main outcome of this study is the identification of word overlap, structural representations, and MT measures as feature subsets that lead to the best performance results for support vector machines in both paraphrase identification and plagiarism detection on corpora. The performance results achieved by deep learning techniques highlight that these techniques are the most promising research direction in this field.

The recognition rate of computer vision algorithms is highly dependent on the image quality. To enhance the visual quality of the images captured under high-dynamic range (HDR) scenes, we propose an efficient and adaptive tone mapping algorithm based on guided image filter (GIF). The HDR image is compressed adaptively according to its average luminance. Then we decompose it into a base layer and a detail layer using the guided image filter. We improve the base layer and enhance the detail layer simultaneously, and combine the two layers to get the final low-dynamic range (LDR) image. Since the parameters are linked with image statistics, they adaptively fit to various kinds of images. The objective evaluation results on HDR image sets demonstrate the superiority of our proposed algorithm. Meanwhile, the result of our algorithm can reduce the halo artifacts and preserve more detail by subjective observation.

Information retrieval in the cloud is common and convenient. Nevertheless, privacy concerns should not be ignored as the cloud is not fully trustable. Fully Homomorphic Encryption (FHE) allows arbitrary operations to be performed on encrypted data, where the decryption of the result of ciphertext operation equals that of the corresponding plaintext operation. Thus, FHE schemes can be utilized for private information retrieval (PIR) on encrypted data. In the FHE scheme proposed by Ducas and Micciancio (DM), only a single homomorphic NOT AND (NAND) operation is allowed between consecutive ciphertext refreshings. Aiming at this problem, an improved FHE scheme is proposed for efficient PIR where homomorphic additions and multiplications are based on linear operations on ciphertext vectors. Theoretical analysis shows that when compared with the DM scheme, the proposed scheme allows multiple homomorphic additions and a single homomorphic multiplication to be performed. The number of allowed homomorphic additions is determined by the ratio of the ciphertext modulus to the upper bound of initial ciphertext noise. Moreover, simulation results show that the proposed scheme is significantly faster than the DM scheme in the homomorphic evaluation for a series of algorithms.

Person re-identification is one of the indispensable elements for visual surveillance. It assigns consistent labeling for the same person within the field of view of the same camera or even across multiple cameras. While handcrafted feature extraction is certainly one way of approaching this problem, in many cases, these features are becoming more and more complex. Besides, training a deep convolutional neural network (CNN) from scratch is difficult because it requires a large amount of labeled training data and a great deal of expertise to ensure proper convergence. This paper explores the following three main strategies for solving the person re-identification problem: (i) using handcrafted features, (ii) using transfer learning based on a pre-trained deep CNN (trained for object categorization) and (iii) training a deep CNN from scratch. Our experiments consistently demonstrated that: (1) The handcrafted features may still have favorable characteristics and benefits especially in cases where the learning database is not sufficient to train a deep network. (2) A fully trained Siamese CNN outperforms handcrafted approaches and the combination of pre-trained CNN with different re-identification processes. (3) Moreover, our experiments demonstrated that pre-trained features and handcrafted features perform equally well. These experiments have also revealed the most discriminative parts in the human body.

A high resolution microscope is designed for plasma hard X-ray (10–20keV) imaging diagnosis. This system consists of two toroidal mirrors, which are nearly parallel, with an angle twice that of the grazing incidence angle and a plane mirror for spectral selection and correction of optical axis offset. The imaging characteristics of single toroidal mirror and double mirrors are analyzed in detail by the optical path function. The optical design, parameter optimization, image quality simulation and analysis of the microscope are carried out. The optimized hard X-ray microscope has a resolution better than 5$\mu$m at 1mm object field of view. The experimental data shows that the variation of the resolution is smaller in the direction of incident angle decrease than that in the increasing direction.

With the rapid development of smart devices and WiFi networks, WiFi-based indoor localization is becoming increasingly important in location-based services. Among various localization techniques, the fingerprint-based method has attracted much interest due to its high accuracy and low equipment requirement. Traditional fingerprint-based indoor localization systems mostly obtain positioning by measuring the received signal strength indicator (RSSI). However, the RSSI is affected by environmental influences, thereby limiting the precision of positioning. Therefore, we propose a new indoor fingerprint localization system based on channel state information (CSI). We adopt a novel method, in which the amplitude and phase of the CSI are fused to generate fingerprints in the training phase and apply a weighted $k$-nearest neighbor (KNN) algorithm for fingerprint matching during the estimation phase. The system is validated in an exhibition hall and laboratory and we also compare the results of the proposed system with those of two CSI-based and an RSSI-based fingerprint localization systems. The results show that the proposed system achieves a minimum mean distance error of 0.85m in the exhibition hall and 1.28m in the laboratory, outperforming the other systems.

Learning ancestor graph is a typical NP-hard problem. We consider the problem to represent a Markov equivalence class of ancestral graphs with a compact representation. Firstly, the minimal essential graph is defined to represent the equivalent class of maximal ancestral graphs with the minimum number of invariant arrowheads. Then, an algorithm is proposed to learn the minimal essential graph of ancestral graphs based on the detection of minimal collider paths. It is the first algorithm to use necessary and sufficient conditions for Markov equivalence as a base to seek essential graphs. Finally, a set of orientation rules is presented to orient edge marks of a minimal essential graph. Theory analysis shows our algorithm is sound, and complete in the sense of recognizing all minimal collider paths in a given ancestral graph. And the experiment results show we can discover all invariant marks by these orientation rules.

This paper proposes a computationally inexpensive algorithm that utilizes player data to optimize nonplayer character pathfinding in a competitive, multiplayer environment and focuses on imitating the player behavior. The algorithm’s input consists of player statistics gathered during the current and previous matches with additional time and space context, similar in design to influence maps. The input is then enriched with two additional, novel variables, allowing easy online fine-tuning of the output. The obtained result influences the final edge values of the map graph. Any known pathfinding algorithm that works with digraphs can then be utilized to control the agent. This paper contains exemplary results obtained when analyzing input on a map modeled after an existing map in the video game Unreal Tournament.

Currently, there are four drawbacks for filled function methods: (1) too many local optimal solutions result in huge difficulty to search global optimal solutions; (2) difficult to control the parameter(s); (3) difficult to determine the initial point for minimization of filled function; (4) the shallow basins will affect the solution precision during the minimization of the filled function. To overcome these drawbacks, in this paper, we adopt a flatten function to eliminate many local optimal solutions first, and then a new filled function with one parameter is proposed, and its parameter is easy to control. Furthermore, we propose an efficient method for determining initial point of the filled function by using adaptive step size. Moreover, when some basins of the filled function are shallow ones, it will result in inefficiency of searching these basins during the minimization of the filled function. To tackle this issue and make the search for global optimal solutions much easier, we propose a strategy of basin deepening. By integrating these schemes, we propose a new efficient filled flatten function method. Numerical results indicate the efficiency and effectiveness of the proposed filled function methods.

Process discovery is a technique for obtaining process model based on traces recorded in the event log. Nowadays, information systems produce streaming event logs to record their huge processes. The truncated streaming event log is a big issue in process discovery because it inflicts incomplete traces that make process discovery depict wrong processes in a process model. Earlier research suggested several methods for recovering the truncated streaming event log and none of them utilized Coupled Hidden Markov Model. This research proposes a method that combines Coupled Hidden Markov Model with Double States and the Modification of Viterbi–Backward method for recovering the truncated streaming event log. The first layer of states contains the transition probability of activities. The second layer of states uses patterns for detecting traces which have a low appearance in the event log. The experiment results showed that the proposed method recovered appropriately the truncated streaming event log. These results also have proven that the accuracies of recovered traces obtained by the proposed method are higher than those obtained by the Hidden Markov Model and the Coupled Hidden Markov Model.