Search

To overcome the problem of data transmission of the aeroengine health management system, a multilink communication system combining ultrahigh-frequency communication link, 4G cellular mobile communication link, and BeiDou satellite communication link was proposed. This system can realize the functions such as data receiving and sending, data encryption, and resuming transfer from the break point based on multiple links. When the flight altitude is not high, the communication distance is short, so the UHF digital transmission radio communication link is adopted, which is highly efficient and stable. When the communication distance is long, the 4G cellular mobile communication link can ensure both the communication distance and the communication rate. In the area where 4G signal cannot be covered in extreme terrain environment, BeiDou satellite communication link is used for data transmission. Besides, in order to ensure the communication rate of the link, a multilink adaptive switching technology was also developed. The test verified that the system can perform adaptive switching among multiple links, realize air-ground data communication in the whole airspace, and achieve a good communication rate, which has significative value of engineering application.

In order to reduce parking difficulties caused by small parking spaces, low driver driving experience, and complex parking environment, parking assistance systems have attracted attention, but the identification of parking spaces is a key problem and technical difficulty. In this paper, an improved parking space recognition algorithm based on panoramic vision is proposed. Firstly, in the look-around image forming part, a method of distortion correction (DC) and perspective transformation (PT) based on LUT (Look Up Table) transformation is proposed to improve the processing speed of the algorithm. Then, to improve the accuracy of parking space recognition, an improved method combining rough extraction and fine matching is proposed to identify parking spaces in a look-around image. The experimental results show that the method achieves a detection rate of 97.63% under sufficient illumination and 79.77% even under insufficient illumination.

As a popular technology in the field of artificial intelligence, computer vision is gradually adapting to the needs of convenience for human beings, improving production efficiency and reducing production costs. Therefore, this study proposes a computer vision algorithm to locate and identify the location of defects. For the traditional edge detection algorithm Sobel, LoG, Canny, the decisive factor for the detection effect of paint defect image is the adjustment of parameters, which can’t achieve an adaptive edge detection algorithm for paint defects, so it is thought that the evolution idea of ant colony algorithm can be used to achieve accurate detection of defects. This paper proposes an automatic detection method for vehicle body paint film defects based on computer vision. An ant colony optimization edge detection algorithm based on automotive paint features (APF-ACO) is proposed. By combining global update and local update, the convergence speed of ant colony algorithm is improved and a new pheromone calculation and update method is proposed to effectively preserve the edge details of the detected image. A reflection area detection algorithm based on HSV color space is designed to detect the reflective area and eliminate interference. Establish defect classification identification rules, identify and mark five types of defects, and determine defect categories. Experiments show that the method can effectively detect the defect area and the recognition accuracy is 97.76%.

In order to take advantage of the powerful advantages of heterogeneous devices and improve the robustness of the vehicle-mounted surround vision algorithm(VSVA), several key technologies of VSVA are improved in the paper. Firstly, computationally intensive tasks are calculated by heterogeneous Graphics Processing Unit(GPU), at the same time, so as to adapt to the VSVA, the memory model and computing model of GPU are optimized. Then a perspective transformation algorithm based on geometric constraints is proposed to improve the quality of the transformed image. Finally, an image alignment and fusion algorithm based on a calibration board is proposed, which reduces the complexity of the algorithm while ensuring the robustness of the image fusion algorithm. The paper compares the proposed algorithm with the traditional algorithm, the test results show that the proposed algorithm has good robustness and the overall performance of the VSVA is improved to 95.39%, the proposed algorithm can be widely used.

In order to improve the efficiency and accuracy of manual vehicle paint defect detection, the computer vision technology and deep learning methods is used to achieve automatic detection of vehicle paint defects based on small samples in this study. The vehicle body paint defect image was collected in real time, and a new data enhancement algorithm was proposed to enhance the database for the over-fitting phenomenon caused by small sample data. Aiming at the defect characteristics inherent in vehicle paints, an improved MobileNet-SSD algorithm for automatic detection of paint defects is proposed by improving the feature layer of MobileNet-SSD network and optimizing the matching strategy of bounding box. The experimental results show that the improved MobileNet-SSD algorithm can detect the defects of six traditional body paint films with an accuracy rate of over 95%, which is 10% faster than the traditional SSD algorithm, and can realize real-time and accurate detection of body paint defects.

The traditional pedestrian navigation system that uses zero velocity update algorithm cannot calculate traveled distance accurately or observe the heading error. A new model called body odometer (BOR) that consists of a step length model and a correction factor is proposed to obtain precise single step length for dead reckoning. A BOR/MINS integrated framework uses the difference between micro electro mechanical inertial navigation system (MINS) calculated distance and single step length, as a new observation to estimate the correction factor and compensate navigation errors via a Kalman filter. To eliminate the heading error accumulation, a new gyro drift reduction method that combines heuristic drift reduction method and complementary filter is presented. The 200 m straight line experiments show that the calculated distance by BOR/MINS integrated method is much closer to the real distance with the average error percentage of 0.24%. Three differently designed trajectories’ experiments show that the proposed method has a higher match degree with the real trajectories and the positioning error with respect to the total traveled distance is less than 0.6%.

Due to the issue of disorder, it is difficult to directly utilise a 2D convolutional neural networks to process 3D point clouds. Recently, PointNet can directly use 3D point sets as the input of convolutional neural networks and complete the processing of point clouds with multi-layer perceptron (MLP) and symmetric functions. However, the use of MLP to compute the weight function ignores the problem of non-uniformity sampling caused by the density of point set data. To address the above problem, based on the PointNet++ structure, a kernel density estimation based method is proposed to calculate the density level of the local point sets region under the optimal bandwidth selection principle, and the density re-weighting of the weight function is developed to better fit the structure of local point clouds. In addition, the authors utilise the upsampling convolution operation to avoid duplicate storages and calculations, making the point cloud reconstruction more efficient. The experiments carried out the semantic segmentation on both the synthetic data and the real indoor scenes show that the proposed method is capable of obtaining promising semantic segmentation results.

In recent years, with the large-scale application of distributed power sources in the power grid, the power grid is moving toward low inertia and low damping. The virtual synchronous generator (VSG) has become a hot topic for scholars since it can simulate the moment of inertia, damping and frequency modulation of synchronous generators. Due to its unbalanced load distribution and the random variation of power loads, grid voltage is asymmetrical. Under grid voltage asymmetry, a VSG experiences output current imbalance and power oscillation. The grid current imbalance further reduces the transformer’s three-phase voltage, which makes it unequal, cyclic, and very easy to cause electricity accidents. Aiming at this problem, this paper proposes three modes of current balance, reactive power balance and active power balance without changing the characteristics of VSGs. The output current balance and power are constant when the grid voltage is asymmetrical. The effectiveness and feasibility of the control strategy are verified by simulation and experimental results, which provides an effective scheme for balanced and stable operation of the grid.

The power-split hybrid electric vehicle achieves excellent fuel economy because both the engine speed and the torque of this system are decoupled from the road load. However, for a power-split hybrid electric vehicle with multiple power sources, the inconsistency of the response characteristic of each power source seriously affects the stability control of the power system and riding comfort, so the coordinated control of the power system is particularly important. This article proposed a dynamic coordinated control strategy. First, extended Kalman filter is applied to realize robust online estimation of the engine dynamics. Then, an extended Kalman filter–based and model predictive control–based dynamic coordinated control strategy is designed to achieve accurate reference tracking in hybrid electric mode. Considering the real-time performance for the online application of the dynamic coordinated control strategy, a fast model predictive control solver is formed based on a reasonable assumption. Offline simulation results show that accurate reference tracking is achieved in hybrid electric mode. Hardware-in-the-loop simulation is also conducted to validate the real-time performance of the proposed dynamic coordinated control strategy. This study is expected to improve the performance and robustness of the dynamic coordinated control strategy in hybrid electric mode while reducing the calibration load.

In this paper, an automated edge detection model is developed for large irregular circular using a laser scanning approach. Some key influencing factors, just like surface quality, surface orientation and scan depth, have been respectively considered in this model. By consider of the limited of field of view and corresponding optical constraints, only a small part of key region is effectively captured by laser scanner at a specific posture. In term of specified view angles of laser scanner, the position of laser scanner approximate to the stand-off distance with respect to edge area can be determined and then high-quality point clouds of large irregular circular edge feature can be effectively obtained. Finally, a series of experiments are performed on workpieces with different large irregular circular features. The experimental result shows our method features high automation and high efficiency. Those results are most promising for on-machine applications in dimensional measurement of large-scale workpiece.

Infrared images have the advantages of clear object outlines and rich thermal information, but the vision is blurred and detailed information is lacking. Visible light images are the opposite. The two are complementary, and more valuable images can be obtained through fusion, making subsequent judgments easier. This paper proposes a fusion method based on jump connection convolution layer aiming at the problem of the low utilization rate of feature map and bad restoration effect of the original image of current fusion methods based on deep learning. By designing the network structure and loss function, this method guides the generated result graph to move closer to the source image with more gradient information. Experiments show that the fusion network can generate ideal results and has obvious advantages in the restoration of the source image.

Trajectory control of tunnel boring machine (TBM) has important implications for excavation efficiency and tunnel quality. A novel rectification trajectory planning method is proposed for TBM with adaptively designed direction and curvature against different attitude deviations and target path, and is comparatively studied with traditional method via numerical simulations. TBM could be fully-actuated by decoupling thrust and torque cylinders in horizontal and vertical planes, respectively. After the tropology analysis of the 4-SPS/PS structure, the real-time expected motions of torque cylinders in vertical plane are derived via reverse kinematics together with synchronous thrust cylinders; so does cylinders in horizontal plane. The proportional direction valves of torque and gripper cylinders are compensated with fixed pressure drop. A compound displacement tracking controller could be established, including flow-speed feed-forward with dead-band compensation and displacement feedback by fuzzy proportional–integral (PI) controller with separated integration. Synchronous controller of integral separated PI structure is also proposed. Experiment results on a Φ 2.5 m scaled TBM indicate that, the displacement tracking performance of cylinders under the compound controller against unbalanced load, which is ± 0.9 mm for high-speed thrust cylinders and ± 0.13 mm for low-speed gripper and torque cylinders, could meet the tolerance of trajectory planning and realize accurate attitude correction.

In order to reduce the traffic accidents caused by the blind area, vehicle panoramic view system has been paid more and more attention. However, the panoramic system is a complex and difficult system. In this paper, we propose an improved vehicle panoramic image generation algorithm. Several key technologies have been improved to ensure reliability and efficiency. First of all, we improve the spherical perspective projection algorithm (SPP) based on the scanning line idea and bilinear interpolation to rectification the fisheye image. Then the inverse perspective projection mapping of undistorted image is used to obtain a top view. In order to reduce computation, the method of manually selecting the target point is carried out. Finally, SURF algorithm is used to find the feature points between the bird’s-eye view images around vehicle. We further put forward to utilize a RANSAC algorithm based on block matching to eliminate the mismatched points in the key point matching process. Experimental results indicate that our vehicle panoramic image generation method works efficiently. The proposed algorithm can effectively remove the serious distortion of fisheye lens, and generate a panoramic image around the vehicle in the end. It possesses good robustness, and can be widely used.

Extracting trusted label from unlabeled data and enhancing the discriminative ability of features are essential issues in person re-identification. The latest study shows that the progressive unsupervised learning has a good performance, and its advantage is specially shown in the ability to select reliable samples from the unlabeled dataset by clustering pedestrian features. Since the features extracted from the same pedestrian samples by the convolutional neural network trained by a softmax loss function may not have good clustering properties, a centralized clustering loss function (CCLF) is proposed in the letter. During the training process, CCLF makes the features of the same category close to the clustering center selected by $k$ -means++, and implements it by adding a distance penalty term on softmax. In the evaluation stage, the study finds that the features have better clustering characteristics and the discrimination is enhanced. The experiments were evaluated on datasets CHUK03, Market1501, and DukeMTMC-reID to verify the superiority of CCLF.

The switching between a damped and an undamped Inertial Navigation System (INS) is an important technical method to ensure its long-term accuracy. The stability of switching is of great importance. This paper studies the switching stability problem between a damped and an undamped INS. A model of an inertial navigation switching system is established by introducing switched control. The average dwell time method is used to analyse stability and a sufficient condition of exponential stability is given. The condition is also extended to the switched system containing constant disturbance and the sufficient condition of exponential stability. The effect of introducing switched control for the smooth operation of the system is verified and the accuracy of a long-term INS is improved effectively.

The presence of operational satellites or small-body space debris is a challenge for autonomous ground-based space object observation. Although most space objects exceeding 10 cm in diameter have been cataloged, the position of each space object (based on six orbital parameters) remains important and should be updated periodically, as the Earth’s orbital perturbations cause disturbances. Modern ground-based passive optical telescopes equipped with complementary metal-oxide semiconductors have become widely used in astrometry engineering, being combined with image processing techniques for target signal enhancement. However, the detection and tracking performance of this equipment when employed with image processing techniques primarily depends on the size and brightness of the space target, which appears on the monitor screen under variable background interference conditions. A small and dim target has a highly sensitive tracking error compared to a bright target. Moreover, most image processing techniques for target signal enhancement require large computational power and memory; therefore, automatic tracking of a space target is difficult. The present work investigates autonomous space target detection and tracking to achieve high-sensitivity detection and improved tracking ability for non-Gaussian and dynamic backgrounds with a simple system mechanism and computational efficiency. We develop an improved particle filter (PF) using the ensemble Kalman filter (KF) for track-before-detect (TBD) frameworks, by modifying and optimizing the computational formula for our non-linear measurement function. We call this extended version the “ensemble Kalman PF-TBD (EnKPF-TBD).” Three sequential astronomical image datasets taken by the Asia-Pacific Ground-Based Optical Space Objects Observation System (APOSOS) telescope under different conditions are used to evaluate three proposed TBD baseline frameworks. Given an optimal random sample size, the EnKPF-TBD exhibits superior performance to PF-TBD and threshold-based unscented KF with two-dimensional peak search (2dPS). The EnKPF-TBD scheme achieves satisfactory performance for all variable background interference conditions, especially for a small and dim space target, in terms of tracking accuracy and computational efficiency.

Scientific research data has an important strategic position for the development of enterprises and countries, and is an important basis for management to conduct strategic research and decision-making. Compared with the Internet industry, big data technology started late in the military enterprises, while military enterprises research data often has the characteristics of decentralization, low relevance, and diverse data types. It cannot fully utilize the advantages of data resources to enhance the core competitiveness of enterprises. To this end, this paper deeply explores the application methods of big data technology in military scientific research data management, and lays a foundation for the construction of scientific research big data platform.

A multi-factor reputation evaluation model of Internet of vehicles is proposed according to the malicious vehicle node and false information in the IOV (Internet of Vehicles). The model comprehensively considers the influence of traffics, information, environment and other factors on vehicle node reputation, and establishes a model of vehicle reputation evaluation. Facing with the characteristics of information of multiple applications, quantifies the influence of different factors and types of information on vehicle reputation.

In view of flameproof electrical apparatus gap measurement demand in the current coal mine and other flammable and combustible environment, this paper proposed a new automatic measurement method for mine flameproof clearance measuring, which is fast and easy to operate and satisfy the demand of gap measurement of bulk flameproof equipment under mine. Based on the image processing method, the measuring instrument uses the microscope camera to collect the image, and the gap width is obtained by analyzing and processing the gap in the image. The measurement is fast, real-time and portable, which can significantly improve the efficiency of the current underground safety inspection.

In the integrated operation mode of the smart grid dispatching and control system, there is a large amount of real-time data such as important alarms and manual operations that need to be shared between systems. How to synchronize these real-time data across systems with high reliability and performance becomes a problem to be solved. This paper studies the application scenarios of cross-system message synchronization and proposes a scheme based on redundant dual networks, including hardware and software architecture. The key technologies and important functions were implemented, reliability and throughput verifications were performed, and the expected results were achieved.

A CMM with multiple probing systems have the power to deliver tremendous benefits to most notably manufacturing, and have the advantage of high automation, high integration and high precision. According to the ISO standard 10360-9, multiple probing system location error should be identify and calibrated before use. In this paper, a location error self-calibration model was established based on a composite artifact. The location error of individual probing system can be respectively determined reference to the position of the probe configuration. An error separation procedure was introduced to correct this location error. A series of representative experiments were performed on a commercial combined probing system produced by our partner. The experimental results show that multiple probing systems location error was effectively reduced from 4.5μm to2.8μm. Also, this calibration evaluation is very apparently practical outside a laboratory due to its simple, portable, low-cost, and rational procedure.

For uplink massive MIMO systems with hundreds of antennas at the base station, the Linear Minimum Mean Square Error (MMSE) signal detection algorithm is near optimal but involves matrix inversion with high complexity. In this paper, we proposed a low complexity detection algorithm in uplink large-scale MIMO based on Reactive Tabu Search (RTS) algorithm by using SOR iterative algorithm as the initial solution vector algorithm. The simulation result shows that it can reduce the computational complexity from ο(K3) to ο(K2), where K is the number of users. Under the premise of BER performance of the original algorithm, the simulation result shows that the performance of SOR-RTS method is always close to the original RTS algorithm.

Parallel connection of MOSFET devices is an available solution for low-voltage and high-current field. But the current sharing among paralleled power MOSFETs is hardly realized, on account of the parameter variations of MOSFETs, the uneven scattered parameters of layouts or packages, the different parameters of gate drive and so on. The paper analyzes the characteristic parameters of MOSFET and effects of circuit parameters on the static and dynamic drain current. In this paper, the current balance of each parallel branch is realized by adopting the current-sharing method of coupling inductor that coupling coils of common magnetic core are connected in each branch of parallel connection, according to Faraday’s law of electromagnetic induction and the principle of flux constraint. Then, the circuit and mathematical models of coupling inductors were presented to reveal its mechanism to eliminate the unbalance current actively. Finally, the effectiveness and feasibility of the current-sharing method for parallel power MOSFET with series-coupled inductors are verified by simulation.

With the increasing scale of the power grid, there exists more and more equipment in it. Thus, the probability of accidents due to the fault of a certain equipment is getting higher. Therefore, it is significant to detect and diagnose the abnormal equipment timely and effectively to keep power grid safety and steady. In this paper, we propose a fault diagnosis system to address this critical problem. Specifically, the system uses the Supervisory Control and Data Acquisition (SCADA) module to collect the switch quantity information, and conducts single fault diagnosis based on Bayesian network. In addition, it also adapts the fault recorder to obtain the electricity quantity information, and performs multiple fault diagnosis based on D-S evidence theory and fuzzy C-Means (FCM) algorithm. Ultimately, the results demonstrate that the proposed diagnosis system has high accuracy and practicability.

False data injection attacks (FDIAs) are new attack manners for power system state estimation in smart grid. FDIAs often bypass the monitoring and defense of the power system, and the security of the whole system will be compromised after the wrong system running state is transferred to the dispatching center. To address this issue, we put forward a new FDIAs detection system. Firstly, the system preliminarily filters the power system measurements through using the trust white list algorithm. Next, to further improve the abnormal database, self-adaptive particle swarm optimization-back propagation (AMPSO-BP) neural networks are utilized to screen the measurements again. Finally, FDIAs are detected relying on the improved database. Simulation tests demonstrate the accuracy of the proposed detection system.

The cutterhead driving system of tunnel machine is over-actuated by redundant driving chains with inevitable load and parameter deviations. These deviations were rarely considered, and the researches on multi-motors synchronization and gear dynamics were usually isolated. In this paper, a novel electromechanical coupled model of the whole cutterhead driving system is established by connecting the vector-controlled motors via gear meshing system. Force transmissions between all coupled elements are investigated by analyzing rotational and translational dynamics, and the time-varying meshing stiffness and nonlinear backlash of both parallel pinions driving and multi-stage planetary reducer are considered. The phase and frequency features of meshing dynamics are also investigated. Comparative simulations are carried out with load and parameter deviations between multi-chains under two control structures, speed parallel control and torque master-slave control; some practical control principles are also concluded. The simulation results are verified and applied on a Φ2.5-m test rig. The consistent results indicate that, the meshing dynamics have the frequency components of carrier revolution and the meshing frequencies of both current planet and endmost pinion. Torque master-slave control could realize torque synchronization against all the adverse deviations while speed parallel control fails, and the vibration could be significantly reduced with lower rigidity in speed control.

It is a challenging task to use deep learning methods to understand point cloud data and assign semantics due to the complexity of point cloud data structure. In this Letter, a fully convolutional network is designed by the authors to perform point cloud semantic segmentation. The proposed network based on PointNet++ takes point cloud data as input and predicts a semantic label for each point. Their network consists of three parts. In the first, different scale features are extracted, the second part reduces the extracted features and then fuses them, and in the third part, more structural information of the point cloud is preserved by up-sampling with deconvolution to reconstruct the point cloud. They have carried out part segmentation and semantic segmentation of ShapeNet and S3DIS datasets, respectively, and the validity of the network has been verified.

The gait of Amur tiger was studied through video images, and a more accurate individual recognition system of Wild Amur tiger was given on the premise of supplementing pattern recognition technology. Through further research on the common gait of Amur tiger, the basic information database was established to realize the physiological state research of Amur tiger. The characteristic structure and movement standard model of Amur tiger were constructed to complete the simulation and reconstruction of Amur tiger’s routine movement. Through the simulation corridor of tiger movement trajectory, the ecological protection area can be divided effectively.

For the primary mirror of a large-scale telescope, an electro-hydraulic position control system (EHPCS) is used in the primary mirror support system. The EHPCS helps the telescope improve imaging quality and requires a micron-level position control capability with a high convergence rate, high tracking accuracy, and stability over a wide mirror cell rotation region. In addition, the EHPCS parameters vary across different working conditions, thus rendering the system nonlinear. In this paper, we propose a robust closed-loop design for the position control system in a primary hydraulic support system. The control system is synthesized based on quantitative feedback theory. The parameter bounds are defined by system modeling and identified using the frequency response method. The proposed controller design achieves robust stability and a reference tracking performance by loop shaping in the frequency domain. Experiment results are included from the test rig for the primary mirror support system, showing the effectiveness of the proposed control design.

Aiming at the low accuracy of short-term power load forecasting, Short-Term power Load Forecasting Model Based on Fuzzy Neural Network using Improved Decision Tree is proposed. High accuracy power load forecasting for a specific area, specific date and specific time period in the smart grid environment can be implemented for power dispatching. The short-term power load forecasting model preprocesses the historical data of each region and quantifies each feature, then classifies the training data using DBSCAN clustering algorithm and decision tree algorithm MID3, and finally sends the classified data to the fuzzy neural network for training and prediction. Based on the original fuzzy neural network, the short-term power load forecasting model introduces the decision tree to classify the historical data. The simulation results show that the short-term power load forecasting model improves the prediction accuracy, reduces the relative error, and the model is more effective.

A multi-branch parallel simulation method and an architecture of application system are proposed in this paper based on symbiotic potential, and method of construction and calibration of parallel battlefield based on real data, as well as generation of simulation branch based on situation interpretation etc. are proposed. Fast building and adjustment of situation of real time simulation are realized based on simulation to meet the requirements of consistency and parallel operation between virtual simulation and real systems. Multibranch operation management, multi-branch highperformance operation support and other methods are proposed to support the dynamic generation of massive parallel simulation copies based on real-time intelligence and ultra-realtime and large-sample operation to meet the requirements of advanced evaluation of real-time situation. At the same time, a set of simulation application verification frameworks for typical task situation is presented to support the construction of multibranch simulation operation environment, and provide technical verification means for real-time situation prediction and evaluation.

In recent years, countries around the world have been vigorously developing new energy and electrifying new energy sources. As a result an increasing number of power electronic devices appear in the grid which makes the power grid develop toward low inertia and low damping. The Virtual Synchronous Generator (VSG) control strategy can simulate the working state of the synchronous generator and provide the grid with inertia and damping, which has become a research hotspot for scholars. Due to the unbalanced load distribution and the random change of power load, the grid voltage may become unbalanced during the actual operation process. The three-phase output current of the VSG would be unbalanced and a fluctuation of output power appears. To solve this problem, a new current and power quality control strategy based on VSG is proposed in this paper, in which the current balancing or constant power is achieved without changing the basic characteristics of the inverters using VSG control strategy.

The cooperative parking robot completes the parking work in the form of a group of four robots working together. In this paper, we do calculation on some significant parameters which can be used to choose power source. We calculate the torque, speed, and power required by the power source according to the load and speed of parking robot. In addition, we perform force analysis calculations on key mechanisms of the robot. The checking of the strength of key parts and the finite element analysis of the whole structure by using the software ABAQUS are also carried out in this paper. The software ADAMS is also used to analyze forces during the process of lifting and lowering the wheel.

With the continuous advancement of technology, people are paying more and more attention to the living environment of wild animals. Located at the top of the food chain of the forest ecosystem, the Siberian tiger is a flagship species and umbrella species for biodiversity protection. The effective protection and sustainable survival of its population has far-reaching natural and social significance for China’s ecological security1. At present, the research on the Siberian tiger is mainly concentrated in the biological fields such as genes and heredity. Only a small amount of research on the detection of Siberian tigers and their footprints in static images. Research on the behavior recognition of Siberian tigers in videos is currently in its infancy. In the field of computer vision, as an independent research object, more and more researchers pay attention to the automatic recognition of gait. Based on the human gait model, a standard database for evaluating the performance of gait recognition algorithms has been established to promote the research of gait recognition. development of. This article will use deep neural convolutional networks and feature subspace learning algorithms to extract the gait features of the Siberian tiger from video images and establish a gait model of the Siberian tiger. This will lay the foundation for the identification of Siberian tigers and further research on their health status, and provide a foundation for the wild Provide theoretical support for animal protection.

With the continuous infiltration of in-vehicle smart devices, the once-closed in-vehicle network has become an open- ing environment and the problem of in-vehicle information security becomes increasingly prominent day by day. In this paper, we analyzed the harm of in-vehicle information security problem on basis of which we further conducted study on correspond- ing defense demands and summarized the suitable method of analysis; directed against the demand for in-vehicle information security, we conducted analysis on in-vehicle information security defense and summarized the information defending strate- gies of in-vehicle network and in-vehicle field bus; we studied the technologies used in the in-vehicle information security encryption mechanism and made comparative analysis towards the main algorithm.

Fabric defect detection is very important in the textile quality process. Current deep learning algorithms are not effective in detecting tiny and extreme aspect ratio fabric defects. In this paper, we proposed a strong detection method, Priori Anchor Convolutional Neural Network (PRAN-Net), for fabric defect detection to improve the detection and location accuracy of fabric defects and decrease the inspection time. First, we used Feature Pyramid Network (FPN) by selected multi-scale feature maps to reserve more detailed information of tiny defects. Secondly, we proposed a trick to generate sparse priori anchors based on fabric defects ground truth boxes instead of fixed anchors to locate extreme defects more accurately and efficiently. Finally, a classification network is used to classify and refine the position of the fabric defects. The method was validated on two self-made fabric datasets. Experimental results indicate that our method significantly improved the accuracy and efficiency of detecting fabric defects and is more suitable to the automatic fabric defect detection.

To solve the problem of feature sparseness of short texts, we studied the application of LDA Topic Model on feature extension and classification of short texts. Training LDA on external long texts related to short texts, and achieving the inference and extension of short texts’ topics based on LDA solves the feature sparseness of short texts and improves the accuracy of classification effectively. Latent semantic information in LDA can also effectively improve the interpretability of short texts.

The predicting accuracy of geomagnetic field is a major factor influencing magnetic anomaly detection, geomagnetic navigation and geomagnetism. The limitations of current methods consist of complex model, a large number of parameters, method of solving parameters with high complexity and low forecast accuracy during geomagnetic disturbed days. In this paper we explore a deep learning method for forecasting geomagnetic field that adopts structure of recurrent neural networks (RNN) based on long-short term memory (LSTM). This method of LSTM RNN includes analyzing the characteristics of geomagnetic field and training the data set of geomagnetic data with simple and robust mathematical model. Compared with current methods, the high-precision prediction of geomagnetic field based on LSTM RNN is achieved during both geomagnetic quiet and disturbed days. Furthermore, it could be found that the average error and maximum error of LSTM RNN are far smaller than those of the other methods.

Given its highly coupled multi-power sources with diverse dynamic response characteristics, the mode transition process of a power-split Hybrid Electric Vehicle (HEV) can easily lead to unanticipated passenger-felt jerks. Moreover, difficulties in parameter estimation, especially power-source dynamic torque estimation, result in new challenges for jerk reduction. These two aspects entangle with each other and constitute a complicated coupling problem which obstructs the realization of a valid anti-jerk method. In this study, a vehicle dynamics model with reference to a data-driven modeling method is first established, integrating a full-time artificial neural network engine dynamic model that can accurately predict engine dynamic torque. Then the essential reason for the occurrence of vehicle jerks in real driving conditions is analyzed. Finally, to smooth the mode transition process, a more practical anti-jerk strategy based on power-source torque changing rate limitation (TCRL) is proposed. Verification studies indicate that the data-driven vehicle dynamics model has enough accuracy to reflect the vehicle dynamic characteristics, and the proposed TCRL strategy could reduce the vehicle jerk by up to 85.8%, without any sacrifice of vehicle performance. This research provides a feasible method for precise modeling of vehicle dynamics and a reference for improving the riding comfort of hybrid electric vehicles.

Aiming at the problem of maximum power point tracking (MPPT) of PV power generation, combined with the purpose of temperature adjustment of rural PV heating system, the MPPT method based on the model of adjustable load resistance matching is proposed. The output characteristics of PV cells and the principle of load matching are analyzed, and the MPPT of PV generation is realized by perturbation and observation (P&O) method. Regarding the heating equipment as the resistive load, a group of resistors with a wide range and small step length which can be controlled by the relay are designed to test. Finally, taking STM32 as the core processor, adjusting the step size and tracking the change of PV output voltage, the controller can track the change of the external environment accurately, that is, it can track the maximum power point of a PV power system effectively.

To improve the accuracy of sentence similarity computation in automatic question answering system for restricted domain, this paper proposed a new TextRank-RD algorithm based on vector space model. The algorithm assigns the node an initial value based on three factors: whether the domain dictionary contains the word, whether the word is a none, the position of the word. And it uses a weighted graph model that assigns weights according to the importance of nodes rather than an unweighted graph model that assigns weights equally. The experimental results show that the algorithm improved the accuracy of sentence similarity calculation compared with the TF-IDF algorithm based on vector space model, and this has important significance to improve the efficiency of the automatic question answering system for restricted domain.

Online reviews are more and more important for potential consumers to make purchase decisions. The referred information of the new products or unpopular products that have no reviews or very few reviews is limited and this situation makes it difficult for consumers to obtain enough information to understand these products. Indeed, this is a new issue recently that needs to be addressed. In this paper, we study the problem of automatically retrieving meaningful opinion sentences for a new product or unpopular product from reviews of other similar products. The retrieved meaningful opinion sentences should possess three properties: helpfulness, relevance and coverage. We propose a meaningful sentences retrieval model (MSRM), which is centered on these three properties to extract meaningful opinion sentences for a new product or unpopular product. We employ product specifications to estimate similarity between products, and model helpfulness, relevance and coverage properties respectively, finally incorporate these three properties into MSRM to mine meaningful sentences from reviews of similar products. Through a series of experiments on real data sets, experiment results show MSRM achieves much better performance.

This paper studies the combination of information network technologies like Internet of Things (IoT) and big data with traditional Chinese medicine (TCM) to build a system framework oriented to TCM smart service. TCM-oriented knowledge representation technology is also explored so as to realize computer recognition and calculation of TCM health service, the self-learning reasoning technology of system is further studied, and TCM knowledge fuzzy model and modified BP neural network algorithm are introduced into TCM smart service system to conduct machine learning and smart judgment upon various diseases. These technologies will promote the scientific research and artificial intelligence aided diagnosis of TCM.

As a new emerging integrated navigation technology, the GNSS/INS/VO deep integration system can fully leverage the synergistic and complementary characteristics of the three involved subsystems to achieve cost reduction while improving navigation accuracy, availability, and robustness, by mutual aiding between subsystems. In this paper, firstly, based on a brief review of the technical background and research status of the GNSS/INS/VO deep integration technology, the processing flows of this deep integration mode in both signal and information domains are analyzed, and its performance advantages are summarized. Then, the top-level technology roadmap of the GNSS/INS/VO deep integration for UGV navigation in the urban canyon is presented, and several key technical issues from scheme demonstration to engineering implementation are explored. At last, the short and medium term development prospects of the GNSS/INS/VO deep integration are envisaged. The proposed research ideas and system design concept for GNSS/INS/VO deep integration could be used as a reference for relevant technicians in navigation field.

A moving vehicle may very likely run into accidents. The occurrence rate of accidents would be largely reduced if the driver is warned in advance, even only 0.5 s earlier. For a running vehicle, the driving route within short time before collision has the characteristic of Markov. In this case, the coordinates of position only have to be considered within a short range, rather than the running status during the past long period. Within short period before collision, the driving route can be basically divided into two states: a straight line and a binomial curve. In this paper, a mechanism is proposed for sending collision warning messages to running vehicles.

The rapid development of electric vehicles, in addition to strengthening technical research, improve battery life, convenient charging facilities is very necessary. At present, for electric vehicle users, the biggest obstacle to install charging piles in residential parking spaces is from property, and property companies generally refuse to install charging piles for safety reasons. In this paper, a new solution is proposed to replace the original fixed charging pile into movable form. The charging pile is separated from the foundation and connected and fixed with a screw and a lock head. The electric connection adopts the form of air plug, when the charging part is not charged, the small door is opened when charging, and the charging pile is connected with the live charge. The charging pile on the site is relatively low, with the same fixed type, but only when the vehicle in the parking spaces at night charging just installed charging pile body, can reduce the risk of charging pile, the day is just a concrete foundation, and is provided with a lock, so it will not lead to children's shock problems. Small footprint, good safety, so you can reduce the concerns of property companies.