In this paper, Spinning Detail Perceptual Generative Adversarial Networks (SDP-GAN) is proposed for single image de-raining. The proposed method adopts the Generative Adversarial Network (GAN) framework and consists of two following networks: the rain streaks generative network G and the discriminative network D. To reduce the background interference, we propose a rain streaks generative network which not only focuses on the high frequency detail map of rainy image, but also directly reduces the mapping range from input to output. To further improve the perceptual quality of generated images, we modify the perceptual loss by extracting high-level features from discriminative network D, rather than pre-trained networks. Furthermore, we introduce a new training procedure based on the notion of self spinning to improve the final de-raining performance. Extensive experiments on the synthetic and real-world datasets demonstrate that the proposed method achieves significant improvements over the recent state-of-the-art methods.

Unmanned aerial vehicles (UAVs) have been investigated proactively owing to their promising applications. A route planner is key to UAV autonomous task execution. Herein, a hybrid differential evolution (HDE) algorithm is proposed to generate a high-quality and feasible route for fixed-wing UAVs in complex three-dimensional environments. A multiobjective function is designed, and both the route length and risk are optimized. Multiple constraints based on actual situations are considered, including UAV mobility, terrain, forbidden flying areas, and interference area constraints. Inspired by the wolf pack search algorithm, the proposed HDE algorithm combines differential evolution (DE) with an approaching strategy to improve the search capability. Moreover, considering the dynamic properties of fixed-wing UAVs, the quadratic B-spline curve is used for route smoothing. The HDE algorithm is compared with a state-of-the-art UAV route planning algorithm, i.e., the modified wolf pack search algorithm, and the traditional DE algorithm. Several numerical experiments are performed, and the performance comparison of algorithms shows that the HDE algorithm demonstrates better performances in terms of solution quality and constraint-handling ability in complex three-dimensional environments.

Traffic flow prediction is one of the fundamental components in Intelligent Transportation Systems (ITS). Many traffic flow prediction models have been developed, but with limitation of noise sensitivity, which will result in poor generalization. Fused Lasso, also known as total variation denoising, penalizes L₁-norm on the model coefficients and pairwise differences between neighboring coefficients, has been widely used to analyze highly correlated features with a natural order, as is the case with traffic flow. It denoises data by encouraging both sparsity of coefficients and their differences, and estimates the coefficients of highly correlated variables to be equal to each other. However, for traffic data, the same coefficients will lead to overexpression of features, and losing the trend of time series of traffic flow. In this work, we propose a Fused Ridge multi-task learning (FR-MTL) model for multi-road traffic flow prediction. It introduces Fused Ridge for traffic data denoising, imposes penalty on L₂-norm of the coefficients and their differences. The penalty of L₂-norm proportionally shrinks coefficients, and generates smooth coefficient vectors with non-sparsity. It has both capability of trend preservation and denoising. In addition, we jointly consider multi-task learning (MTL) for training shared spatiotemporal information among traffic roads. The experiments on real traffic data show the advantages of the proposed model over other four regularized baseline models, and on traffic data with Gaussian noise and missing data, the FR-MTL model demonstrates potential and promising capability with satisfying accuracy and effectiveness.

In Internet applications, the description for the same point of interest (POI) entity for different location-based services (LBSs) is not completely identical. The POI entity information in a single LBS data source contains incomplete data and exhibits insufficient objectivity. Aligning and consolidating POI entities from various LBSs can provide users with more comprehensive, objective, and authoritative POI information. We herein propose a multi-attribute measurement-based entity alignment method for Internet LBSs to achieve POI entity alignment and data consolidation. This method is based on multi-attribute information (geographical information, text coincidence information, semantic information) of POI entities and is combined with different measurement methods to calculate the similarity of candidate entity pairs. Considering the demand for computational efficiency, the particle swarm optimization algorithm is used to train the model and optimize the weights of multi-attribute measurements. A consolidation strategy is designed for the LBS text data and user rating data from different sources to obtain more comprehensive and objective information. The experimental results show that, compared with other baseline models, the POI alignment method based on multi-attribute measurement performed the best. Using this method, the information of POI entities in multisource LBS can be integrated to serve netizens.

Robotic Manipulators (RM) are nonlinear and coupling system with time-variant and model uncertainties. In addition, RM are subject to different types of disturbances in practice, such as joint frictions, unknown payloads, and interferences from external systems. In this paper, these adverse factors are regarded as disturbance, and classifies them into internal disturbances and external disturbances. In order to achieve high-precision control, a Nonlinear Disturbance Observer (NDO) is designed to suppress external disturbances, and a Fuzzy Logic System (FLS) is designed to compensate internal disturbances. The scheme can effectively suppress the disturbance and improve the control accuracy. The validity of the scheme is shown by computer simulations of a two-link robot manipulator.

Lithium battery packs are the main driving energy source for electric vehicles. A battery pack equalization charging solution using a constant current source for variable rate charging is presented in this paper. The charging system consists of a main constant current source and independent auxiliary constant current sources. Auxiliary constant current sources are controlled by the battery management system (BMS), which can change the current rate of the corresponding single battery, and achieve full charging of each single cell in the series battery pack. At the same time, the state of charge (SOC) is regarded as time series data to establish a long short-term memory recurrent neural network (LSTM-RNN) model, and it is possible to obtain the single battery with lower capacity, so that the charging efficiency and battery pack consistency can be improved. The experimental results show that the open circuit voltage difference between the single cells is less than 50 mV after the charging of 20 strings of lithium battery packs by using this method, which achieve the purpose of equalization charging.

Efficient and precise three-dimensional (3D) measurement is an important issue in the field of machine vision. In this paper, a measurement method for indoor key points is proposed with structured lights and omnidirectional vision system and the system can achieve the wide field of view and accurate results. In this paper, the process of obtaining indoor key points is as follows: Firstly, through the analysis of the system imaging model, an omnidirectional vision system based on structured light is constructed. Secondly, the full convolution neural network is used to estimate the scene for the dataset. Then, according to the geometric relationship between the scenery point and its reference point in structured light, for obtaining the 3D coordinates of the unstructured light point is presented. Finally, combining the full convolution network model and the structured light 3D vision model, the 3D mathematical representation of the key points of the indoor scene frame is completed. The experimental results proved that the proposed method can accurately reconstruct indoor scenes, and the measurement error is about 2%.

In recent years, with the further breakthrough of artificial intelligence theory and technology, as well as the further expansion of the Internet scale, the recognition of human emotions and the necessity for satisfying human psychological needs in future artificial intelligence technology development tendencies have been highlighted, in addition to physical task accomplishment. Musical emotion classification is an important research topic in artificial intelligence. The key premise of realizing music emotion classification is to construct a musical emotion model that conforms to the characteristics of music emotion recognition. Currently, three types of music emotion classification models are available: discrete category, continuous dimensional, and music emotion-specific models. The pleasure-arousal music emotion fuzzy model, which includes a wide range of emotions compared with other models, is selected as the emotional classification system in this study to investigate the influencing factor for musical emotion classification. Two representative emotional attributes, i.e., speed and strength, are used as variables. Based on test experiments involving music and non-music majors combined with questionnaire results, the relationship between music properties and emotional changes under the pleasure-arousal model is revealed quantitatively.

The terminal voltage is easily affected by the characteristics of loads and variations in wind speed, loads and system parameters in a stand-alone wind energy conversion system. This paper presents a terminal voltage control scheme that combines the equivalent-input-disturbance (EID) and model predictive control (MPC). The total disturbance is observed and compensated in real time by the EID. A battery energy storage system based on MPC is employed to smooth the fluctuation and imbalance in power caused by the variation in wind speed and loads, thereby solving the problem of terminal voltage flicker and instability. The appropriate terminal voltage can be obtained using the proposed scheme, which is a simple design and offers good prospects for actual applications. The simulation results demonstrate the validity of the proposed scheme.

To detect the students’ concentration state in classroom, a DS (Dempster–Shafer theory)-based evaluation algorithm is proposed by measuring the students’ Euler angles of their facial attitude. The detection of facial attitude angles can be implemented under the surveillance video with lower pixels. Therefore, compared with other methods for students’ concentration evaluation, the proposed algorithm can be applied directly in most classrooms by the support of existing monitoring equipment. By using DS theory to fuse the concentration state of each student, the curve of students’ overall concentration score changing with time can be obtained to describe the overall classroom concentration state. The design of the algorithm is proved to be feasible and effective under the dataset provided by computer front camera. The realization of the overall function effect of the algorithm is tested under the 35-person classroom video dataset. Compared with the average score from the questionnaire given by 20 reviewers, the accuracy of the proposed algorithm is about 85.3%.

The traditional traffic flow prediction method is based on data modeling, when emergencies occur, it is impossible to accurately analyze the changes in traffic characteristics. This paper proposes a traffic flow prediction model (BAT-GCN) which is based on drivers’ cognition of the road network. Firstly, drivers can judge the capacity of different paths by analyzing the travel time in the road network, which bases on the drivers’ cognition of road network space. Secondly, under the condition that the known road information is obtained, people through game decision-making for different road sections to establish the probability model of path selection; Finally, drivers obtain the probability distribution of different paths in the regional road network and build the prediction model by combining the spatiotemporal directed graph convolution neural network. The experimental results show that the BAT-GCN model reduces the prediction error compared with other baseline models in the peak period.

Based on the dynamic characteristics of electric vehicles, this study describes the use of existing basic parameters of a specific electric vehicle to optimize the performance parameters of an asynchronous motor. In addition, a theoretical reference of such an asynchronous motor is provided.

The fault detection (FD) problem is investigated for event-triggered discrete-time Markov jump systems (MJSs) with hidden-Markov mode observation. A dynamic-event-triggered mechanism, which includes some existing ones as special cases, is proposed to reduce unnecessary data transmissions to save network resources. Mode observation of the MJS by the FD filter (FDF) is governed by a hidden Markov process. By constructing a Markov-mode-dependent Lyapunov function, a sufficient condition in terms of linear matrix inequalities (LMIs) is obtained under which the filtering error system of the FD is stochastically stable with a prescribed H_∞ performance index. The parameters of the FDF are explicitly given when these LMIs have feasible solutions. The effectiveness of the FD method is demonstrated by two numerical examples.

A control algorithm for lateral semi-active suspension based on sliding mode observer is proposed to solve lateral vibration of high-speed trains caused by railway surface excitation. Firstly, the multi-degree-of-freedom vehicle dynamics model of high-speed train is revised on the basis of variation in spring coefficient caused by long-time vibration in practical engineering; Secondly, the observer based on sliding mode variable structure is designed to obtain real-time estimation of unknown interference terms using the equivalent control principle of sliding mode variable structure, which can be further used to calculate the observed values of unknown interference terms; Finally, a control algorithm based on sliding mode observer is proposed. The algorithm inputs complex unknown interference observed values into the sliding mode controller (SMC) as feedback, thereby allowing the observed values to accurately track nonlinear unknown disturbance and weaken the vibration. Stimulation and experimental verification have proven the effectiveness and feasibility of the proposed method.

Multi-target searching is a hotspot and foundation topic in multi-agent systems research. However, most of the research is based on simple environment or known environment, which greatly limits the application of target search. In the non-structured environment, the searching result can be greatly affected by the complex terrain constraints and detectability of targets especially when we have no prior knowledge about the environment. In the paper, a novel search strategy combining maximum visibility and particle swarm optimization is proposed for the target search problem in a completely unknown and non-structural environment. The strategy utilizes the concept of visibility to describe how well the agent detects the map, and guides the agent to perform online path planning to complete the search task. In addition, considering the limited communication distance and communication bandwidth, the strategy introduces a cooperative mechanism for each agent to improve the search efficiency. Finally, in the experimental part, the search strategy is compared with the commonly used search strategies. Compared with the methods combining advantages, the proposed strategy can still achieve similar results, which proves the feasibility and efficiency of the strategy.

The automatic classification of point of interest (POI) function types based on POI name texts and intelligent computing can provide convenience in travel recommendations, map information queries, urban function divisions, and other services. However, POI name texts belong to short texts, which few characters and sparse features. Therefore, it is difficult to guarantee the feature learning ability and classification effect of the model when distinguishing the POI function types. This paper proposes a POI classification method based on feature extension and deep learning to establish a short-text classification model. We utilize an Internet search engine as an external knowledge base to introduce real-time, large-scale text feature information to the original POI text to solve the limitation of sparse POI name text features. The input text information is represented by the attention calculation matrix used to reduce the noise information of the extended text and the word-embedding matrix of the original text. We utilize a convolutional neural network with excellent local feature extraction ability to establish the classification model. Experimental results on a real-world dataset (obtained from Baidu) show the excellent performance of our model in POI classification tasks compared with other baseline models.

We propose a visual servoing (VS) approach with deep learning to perform precise, robust, and real-time six degrees of freedom (6DOF) control of robotic manipulation to ease the extraction of image features and estimate the nonlinear relationship between the two-dimensional image space and the three-dimensional Cartesian space in traditional VS tasks. Owing to the superior learning capabilities of convolutional neural networks (CNNs), autonomous learning to select and extract image features from images and fitting the nonlinear mapping is achieved. A method for designing and generating a dataset from few or one image, by simulating the motion of an eye-in-hand robotic system is described herein. Therefore, network training requiring a large amount of data and difficult data collection occurring in actual situations can be solved. A dataset is utilized to train our VS convolutional neural network. Subsequently, a two-stream network is designed and the corresponding control approach is presented. This method converges robustly with the experimental results, in that the position error is less than 3 mm and the rotation error is less than 2.5° on average.

In this paper, a segmentation method for cell images using Markov random field (MRF) based on a Chinese restaurant process model (CRPM) is proposed. Firstly, we carry out the preprocessing on the cell images, and then we focus on cell image segmentation using MRF based on a CRPM under a maximum a posteriori (MAP) criterion. The CRPM can be used to estimate the number of clusters in advance, adjusting the number of clusters automatically according to the size of the data. Finally, the conditional iteration mode (CIM) method is used to implement the MRF based cell image segmentation process. To validate our proposed method, segmentation experiments are performed on oral mucosal cell images. The segmentation results were compared with other methods, using precision, Dice, and mean square error (MSE) as the objective evaluation criteria. The experimental results show that our method produces accurate cell image segmentation results, and our method can effectively improve segmentation for the nucleus, binuclear cell, and micronucleus cell. This work will play an important role in cell image recognition and analysis.