International Journal of Pattern Recognition and Artificial Intelligence

Streaming time series retrieval (TSR) has been widely concerned in academia and industry. Considering the large volume, high dimensionality and continuous accumulation features of time series, there is limited capability to perform in-depth similarity searching directly on the raw time series data. Therefore, time series representation, which can provide the dimension reduction-based approximate results for the raw data, should be utilized in the first step for streaming TSR. However, the existing representation-based TSR methods mainly have two limitations: on the one hand, the representation efficiency of the current methods is too slow to adapt for real-time streaming time series representation; on the other hand, the retrieval efficiency of them is also not ideal, and thus fails to recognize the specific given sequence patterns on the streaming data effectively. In this paper, we present an efficient retrieval method on streaming time series. Concretely, our method can incrementally represent the features of streaming data to automatically prune the corresponding dissimilar sequences and retain the most similar candidates for efficient one-pass searching. Extensive experiments on real world datasets have been conducted to demonstrate the superiority of our method.

Nowadays, imbalanced data classification is a hot topic in data mining and recently, several valuable researches have been conducted to overcome certain difficulties in the field. Moreover, those approaches, which are based on ensemble classifiers, have achieved reasonable results. Despite the success of these works, there are still many unsolved issues such as disregarding the importance of samples in balancing, determination of proper number of classifiers and optimizing weights of base classifiers in voting stage of ensemble methods. This paper intends to find an admissible solution for these challenges. The solution suggested in this paper applies the support vector data descriptor (SVDD) for sampling both minority and majority classes. After determining the optimal number of base classifiers, the selected samples are utilized to adjust base classifiers. Finally, genetic algorithm optimization is used in order to find the optimum weights of each base classifier in the voting stage. The proposed method is compared with some existing algorithms. The results of experiments confirm its effectiveness.

A feature selection method based on mutual information and support vector machine (SVM) is proposed in order to eliminate redundant feature and improve classification accuracy. First, local correlation between features and overall correlation is calculated by mutual information. The correlation reflects the information inclusion relationship between features, so the features are evaluated and redundant features are eliminated with analyzing the correlation. Subsequently, the concept of mean impact value (MIV) is defined and the influence degree of input variables on output variables for SVM network based on MIV is calculated. The importance weights of the features described with MIV are sorted by descending order. Finally, the SVM classifier is used to implement feature selection according to the classification accuracy of feature combination which takes MIV order of feature as a reference. The simulation experiments are carried out with three standard data sets of UCI, and the results show that this method can not only effectively reduce the feature dimension and high classification accuracy, but also ensure good robustness.

For the logistics sorting warehouse without much light is complex, and the difference between express packaging is not obvious, a fast recognition method of sorting images based on deep learning and dual tree complex wavelet transform was studied. Sorting images are not very clear due to factors such as the enclosed environment and the weak lighting conditions of the warehouse. First, the dual tree complex wavelet transform is used to preprocess the sorting image for noise reduction and other image preprocessing. Second, a convolutional neural network (CNN) was designed. On the basis of Alexnet neural network, parameters of convolutional layer, ReLU layer and pooling layer of CNN are redefined to accelerate the learning speed of neural network. Lastly, according to the new image classification task, the last three layers of the neural network, the full connection layer, the softmax layer and the classification output layer are defined to adapt to the new image recognition. The proposed fast sorting image recognition technology based on deep learning has higher training speed and recognition accuracy in the face of more complicated sorting image recognition, which can meet the experimental requirements. Rapid identification of sorting images is of great significance to improve the efficiency of logistics in unmanned warehouses.

Recognition of hand activities of daily living (hand-ADL) is useful in the areas of human–computer interactions, lifelogging, and healthcare applications. However, developing a reliable human activity recognition (HAR) system for hand-ADL with only a single wearable sensor is still a challenge due to hand movements that are typically transient and sporadic. Approaches based on deep learning methodologies to reduce noise and extract relevant features directly from raw data are becoming more promising for implementing such HAR systems. In this work, we present an ARMA-based deep autoencoder and a deep recurrent network (RNN) using Gated Recurrent Unit (GRU) for recognition of hand-ADL using signals from a single IMU wearable sensor. The integrated ARMA-based autoencoder denoises raw time-series signals of hand activities, such that better representation of human hand activities can be made. Then, our deep RNN-GRU recognizes seven hand-ADL based upon the output of the autoencoder: namely, Open Door, Close Door, Open Refrigerator, Close Refrigerator, Open Drawer, Close Drawer, and Drink from Cup. The proposed methodology using RNN-GRU with autoencoder achieves a mean accuracy of 84.94% and F1-score of 83.05% outperforming conventional classifiers such as RNN-LSTM, BRNN-LSTM, CNN, and Hybrid-RNNs by 4–10% higher in both accuracy and F1-score. The experimental results also showed the use of the autoencoder improves both the accuracy and F1-score of each conventional classifier by 12.8% in RNN-LSTM, 4.37% in BRNN-LSTM, 15.45% CNN, 14.6% Hybrid RNN, and 12.4% for the proposed RNN-GRU.

In order to discover the difference among dual strategies, we propose an alternating primal-dual algorithm (APDA) that can be considered as a general version for minimizing problem which is multiple-summed separable convex but not necessarily smooth. First, the original multiple-summed problem is transformed into two subproblems. Second, one subproblem is solved in the primal space and the other is solved in the dual space. Finally, the alternating direction method is executed between the primal and the dual part. Furthermore, the classical alternating direction method of multipliers (ADMM) is extended to solve the primal subproblem which is also multiple summed, therefore, the extended ADMM can be seen as a parallel method for the original problem. Thanks to the flexibility of APDA, different dual strategies for image restoration are analyzed. Numerical experiments show that the proposed method performs better than some existing algorithms in terms of both speed and accuracy.

With rapid development of intelligent video surveillance systems based on cloud computing devices and edge computing devices in Cyber-Physical-Social Systems, massive surveillance video data has brings enormous challenge for video storage and transmission. However, existing surveillance video coding approaches hardly utilize intelligent video analysis results for improving video coding. This paper proposed a surveillance video coding scheme for traffic scene based on vehicle knowledge and shared library by cloud-edge computing in Cyber-Physical-Social Systems. Firstly, in order to provide the object library for synchronous application at the encode and decode side offline, a generation method of shared long-term foreground reference object library is proposed by using the existing large-scale monitoring vehicle object datasets. Then, to meet the requirement of low complexity and high-performance coding, a virtual foreground reference picture generation method with coding-oriented object retrieval is proposed. Experimental results show that the proposed scheme can obtain the satisfactory effect of the virtual foreground reference picture. Also, it can yield remarkable bit rate reductions, compared to HEVC.

Perceptual video coding (PVC) optimization has been an important video coding technique, which can be consistent with the perception characteristics of the human visual system (HVS). Currently, PVC schemes incorporating the just noticeable distortion (JND) model can obtain better performance gain in all PVC schemes. To further accelerate the JND computation for real-time video coding applications (e.g. surveillance video coding and conference video coding), this paper proposes a fast perceptual surveillance video coding (PSVC) scheme based on background model-driven JND estimation method. First, to utilize the surveillance scene characteristics, the computation complexity of JND estimation can be significantly decreased by reusing the content complexity of background regions. Then we apply the perceptive video coding scheme into the background modeling-based surveillance video codec. The proposed scheme adopts background modeling frame as background anchor. Experimental results show that the proposed scheme can yield remarkable time saving of 42.33% maximum and on average 34.76% with approximate bitrate reductions and similar subjective quality, compared to HEVC and other state-of-the-art schemes.

Pupil parameters are the essential foundation for many applications, such as cognitive science and human–machine interaction. Existing approaches are still affected by various challenges. We propose a novel pupil detection pipeline (known as Edge Points Selector “EPS”) which is suitable even for partial occlusion, lighting, and specular reflection. EPS consists of pupil area detection, edge selection, and ellipse fitting. For the first time, we find the suitable Haar-like feature of 2D-pupil and a new pupil edge feature in the local pupil area, and integrate them into the proposed pipeline. EPS was compared with two state-of-art methods on 130856 images in this work. Within an error threshold of 5 pixels, our method outperforms the comparison algorithms by 33.8% and 19.4%, respectively, on overage.

The traditional fractional-order particle swarm optimization (FOPSO) algorithm depends on the fractional order $\alpha$, and it is easy to fall into local optimum. To overcome these disadvantages, a novel perspective with PID gains tuning procedure is proposed by combining the time factor with FOPSO, i.e. a new fractional-order particle swarm optimization called TFFV-PSO, which reduces the dependence on the fractional order to enhance the ability of particles to escape from local optimums. According to its influence on the performance of the algorithm, the time factor is varied with population diversity parameters to balance the exploration and exploitation capabilities of the particle swarm, so as to adjust the convergence speed of the algorithm, then it follows that a better convergence performance will be obtained. The improved method is tested on several benchmark functions and applied to tune the PID controller parameters. The experimental results and the comparison with previous other methods show that our proposed TFFV-PSO provides an adequate velocity of convergence and a satisfying accuracy, as well as even better robustness.

With the purpose of clarifying the energy-saving influencing factors of the participants in the construction project, qualitative analysis and scientific measurement analysis were combined in this paper to define and identify the energy-saving influencing factors of the construction project lifecycle. First, a qualitative analysis of the research literature of the natural science index (SCI) and social science citation index (SSCI) databases was conducted through a combination of literature analysis and expert interviews, constructing a list of the energy-saving influencing factors in the lifecycle of construction projects. Second, the knowledge map of the energy-saving research literature of construction projects in the past 20 years from the Web of Science (WOS) database was drawn through Citespace. Then the hotspots of energy-saving research in international construction projects were identified through keyword co-occurrence map as energy saving, energy efficiency, and energy consumption. The main research subjects were divided into four categories: energy-saving management and incentive policies, project users’ energy-saving behaviors and awareness, various energy-saving new materials and technologies, and energy-saving management and control at the operation and maintenance stage. Next, nine main clusters on energy-saving research were identified through cluster analysis, mainly distributed in various energy-saving technology research, energy-saving behavior and power research of owners and residents, policy and regulation research, and energy-saving management and control research in the operation and maintenance stage. Finally, the knowledge mapping analysis results were classified and defined according to the action scope of each stage of the construction project, and the identification of the energy-saving influencing factors in the construction project lifecycle was completed.

With the rapid development of sharing economy, bike-sharing becomes essential because of its zero emission, high flexibility and accessibility. The emergence of the public bicycle system not only alleviates the traffic pressure to a certain extent, but also contributes to solving the “last kilometer” problem of public transportation. However, due to the concentrated use of shared bikes, many shared bikes are left in disorder, which seriously affects the urban environment and causes traffic problems. How to manage the allocation of bike-sharing and improve the city’s shared cycling system have become a highly discussed issue. We, taking Beijing as an example, research on the allocation of shared bikes by using the open source data provided by Amap, Baidu Map and websites of shared bikes, which are used to analyze the allocation, and establish an optimizing comprehensive evaluation model to evaluate the required level. In the end, we look forward the future of bike-sharing market.

Motivated by the operation of bulk terminal in the Pearl River Delta, we study a berth allocation problem (BAP) with 2 berths and overall quay length constraint. The quay length occupied by a vessel depends on its berthing directions. The feasible berthing direction changes in a tide cycle. The objective is to minimize the total service of all vessels. We develop a mixed integer programming (MIP) model for the problem. We also propose an efficient genetic algorithm to tackle the problem. Our computational experiments show that the mixed integral programming model can only be solved by CPLEX for small-size instances, the genetic algorithm obtains good approximation solutions for large-size instances.

Balancing the convergence and diversity of many-objective evolutionary algorithms is difficult and challenging. In this work, a multi-search strategy based on decomposition is proposed to generate good offspring and improve convergence, and an external population strategy is used to maintain the diversity of the obtained solutions. The multi-search strategy allows the selection of sparse and convergent nondominated solutions to carry out the exploration and exploitation steps. Experiments are conducted on 15 benchmark functions from the CEC 2018 with 5, 10, and 15 objectives. The results indicate that the proposed algorithm can obtain a set of solutions with better diversity and convergence than the five efficient state-of-the-art algorithms, i.e. NSGAIII, MOEA/D, MOEA/DD, KnEA, and RVEA.

This paper combines the traditional output regulation (OR) theory with fuzzy identification method, and designs a distributed output regulator for the multi-agent system (MAS). In view of the situation that the follower model in the MAS cannot be obtained accurately, the fuzzy model is established to make it approximate to the considered nonlinear discrete follower agent system first. By employing fuzzy identification method and OR theory, a distributed controller is proposed, make follower agents track the reference signal given by the dynamic leader. Two numerical examples demonstrate the obtained results.