Showing papers presented at "International Conference on Measuring Technology and Mechatronics Automation in 2020"

Addressing Crime Situation Forecasting Task with Temporal Graph Convolutional Neural Network Approach

Abstract: Crime situation forecasting has always been a challenging task for smart city security decision support system construction. Accurate crime dynamic capture can optimize the allocation efficiency of police resources to respond to various crime situations. Existing crime forecasting approaches mainly utilize some stochastic process modeling methods, some statistical learning models or some pixel-level deep learning models to capture spatio-temporal dynamic. Compared with traditional models, Graph Neural Network (GNN) obviously can better capture the spatial structural features other than Euclidean distance. In this paper, we address the crime situation forecast task with Temporal Graph Convolutional Neural Network (T-GCN) approach, a graph deep learning approach for spatio-temporal dynamics capturing. Graph Convolutional Neural Network (GCN) and Recurrent Neural Network (RNN) are combined in T-GCN model to capture the spatial and temporal dynamics respectively. In experimental studies, we evaluate T-GCN model on crime statistics dataset of San Francisco City and demonstrate effectiveness of T-GCN compared with some traditional state-of-art baseline models.

Real-Time Detection of Vehicle Speed Based on Video Image

Abstract: With the continuous development and improvement of Intelligent Transportation System, more and more attention has been paid to the real-time and accurate speed information detection. Video vehicle speed detection based on machine vision has attracted the attention of the researchers due to its practical convenience and other advantages. In this paper, we propose a target tracking method based on video image feature matching vehicle to simplify the algorithm complexity and obtain high accuracy. The background subtraction method based on KNN algorithm is used to identify vehicle targets and obtain good initial vehicle characteristics. Experimental results show that the method has high real-time and the relative error of vehicle speed detection can be controlled at about 5%.

Defect Inspection Algorithm of Metal Surface Based on Machine Vision

Abstract: During the production and post-processing of complex metal parts, surface defects such as scratches, stains, and pits will inevitably occur, which will reduce the yield of metal parts and cause serious economic losses. Complex metal surfaces have complex surface texture characteristics, which leads to the difficulty of detection during the detection process. Therefore, this paper presents a defect inspection algorithm of metal surface based on machine vision. The proposed surface defect inspection algorithm first uses improved bi-dimensional empirical mode decomposition (BEMD)-based extracting algorithm to perform initial extracting of surface defects through filtering out complex textures on the metal surface, while retaining as much effective information as defects as possible. Then, the inspection algorithm applies Canny edge detection operator to detect the edge information of these defects. Finally, the final acquisition of defect edges can be achieved by connecting edge breakpoints and image filling operations. Experimental results on inspection of a variety of the surface defects on parts with metal surface are reported to show the performance of the defect inspection algorithm.

An Improved Particle Swarm Optimization Algorithm Based on Entropy and Fitness of Particles

Abstract: Particle swarm optimization (PSO) is an algorithmic technique for optimization by solving a wide range of optimization problem. This paper presents an improved PSO. The proposed algorithm consists of two parts. Firstly, population initialization method based on entropy is proposed. Secondly, an improved accelerated learning coefficient is proposed. In this algorithm, a modified velocity updating formula of the particle is used, where the randomness in the course of updating particle velocity and the acceleration coefficient is relatively decreased. The entropy of each dimension is calculated to decrease the randomness of swarm and increase population diversity. Each particle has a different learning rate according to its fitness during evolution, which can balance the global and local searching ability of the population and avoid falling into local optimum. Experimental results show that, the proposed algorithm remarkably improves the ability of PSO to jump out of the local optima and significantly enhance the convergence precision.

Pso Optimization of Ladrc for the Stabilization of a Quad-Rotor

Abstract: The quad-rotor is an under-actuated, strong coupled nonlinear system with many parameters, and the method of tuning them is very tough. For the sake of the stabilization of a quad-rotor, we propose the use of Particle Swarm Optimization (PSO) algorithm for tuning Linear Active Disturbance Rejection Control (LADRC) which is applied for the stabilization of a quad-rotor model. It is demonstrated that LADRC based on PSO algorithm has better flexibility, adaptability and robustness than that of trial and error method to determine parameters, and can improve the accuracy of the control system.

Study on Noise Elimination of Mechanical Vibration Signal Based on Improved Wavelet

Abstract: The analysis and processing of vibration signal is an important means of mechanical equipment fault diagnosis. However, the collected vibration signal is inaccurate or even wrong because of noise. Therefore, the signal must be denoised. This paper introduces the second generation wavelet principle based on lifting method and the construction method of adaptive redundant second generation wavelet. The effect of the redundant second generation wavelet transform and the adaptive redundant second generation wavelet transform applied to the vibration signal denoising is compared. Experimental results show that the self-adaptive redundant second generation wavelet is better in signal denoising.

Feedback Linearization of MIMO Nonlinear System with Measurable Disturbance

Abstract: A class of MIMO nonlinear system with measurable disturbance is studied in this paper. Based on the relative degree concept of the nonlinear system in differential geometry theory, the relative order vector of measurable disturbance of MIMO nonlinear system is defined. And a nonlinear feedforward / state feedback transformation law is proposed which can eliminate the effect of measurable disturbance on the output of the system largely or completely, transforming the nonlinear system into a linear closed-loop system, providing favorable conditions external controller design. In addition, the application range of the feedback linearization transformation law and the specific requirements of the exact linearization are analyzed in detail in this paper. Finally, simulation examples are given to demonstrate the effectiveness of the transformation law.

Research on the Application of GA Improved Neural Network in the Prediction of Financial Crisis

Abstract: Aiming at the prediction of the financial distress of the company and taking the Chinese manufacturing listed companies as the research object, the abnormal financial situation is specially treated as the sign of the listed companies in financial distress. Genetic algorithm is used as the pre-device of the neural network model to optimize the initial value and threshold value of the network input, shorten the training time of the network, and improve the prediction efficiency of the network. The sample data are from real financial statements of listed companies, and 67 companies with normal and abnormal financial conditions are selected for empirical research. The original financial crisis early warning model and the financial crisis early warning model based on genetic algorithm are constructed. Empirical research shows that the prediction accuracy of the optimization model is higher than that of the traditional BP neural network model, which can improve the financial crisis early warning ability and analysis efficiency of listed companies from the perspective of technology.

A Method to Improve the Fresh Data Query Efficiency of Blockchain

Abstract: In order to accelerate the improvement of the blockchain's efficiency of tracing fresh data, this paper proposes a novel method of tracing data based on LevelDB. In the process of blockchain source tracing, it is necessary to query data in the history database This paper proposes an improved method for querying high-fresh data in memory while consuming a small amount of space on the commonly used database LevelDB in the blockchain. Hash indexes are applied to LevelDB's memory and build hash indexes on MemTable and Immutable MemTable respectively. When MemTable is converted to Immutable MemTable and data is stored from memory to disk, the index will be converted and deleted accordingly to prevent the index from occupying too much space. Indexing the data in memory makes the query speed of fresh data faster and because of the amount of data will not be too large, the serious hash conflicts will not be resulted in. The experimental results show that the query speed for data in memory has been improved with the limition of writing performance impaction.

Twist Angle Compensation for Three Dimensional Real-Time Shape Sensing Method Based on Multi-Core Optical Fiber

Abstract: Multi-core fiber is attractive for shape sensing because of its spatial symmetrical distribution of the different cores. A real-time shape sensing approach to enhance the precision of the three dimensional reconstruction shape based on multi-core optical fiber gratings is proposed and demonstrated. The precision of reshaping curve is improved by correcting fiber twisting angle error resulted from twisting loads which are inevitable when fiber bends. By modeling the bending state of optical fiber under twisting load, the twist angles between neighboring fiber gratings are derived from measured strain signal. To be compared with standard deviation of 49.9°bending angle error, the experimental results show that the error of bending angle is less than 8.9°when twist angle measurements are taken into the shape solution and thus the error of curve points position are less than 2.3% in comparison with actual curve points position and absolute cumulative position error is less than 5 cm for 1-meter-long fiber test.

Performance Evaluation of Faults in a Photovoltaic Array Based on V-I and V-P Characteristic Curve

Abstract: The prediction and forecasting of faults occurring in photovoltaic (PV) system is one of the important aspect to escalate the reliability, output power generation, proficiency, lifetime and effectiveness of overall system. In this article, the behavior of PV array is studied under various faulty circumstances and new approach has offered for the better performance assessment of PV array. This approach is established on the analysis of peculiarities displayed by voltage power (V-P) and voltage-current (V-I) characteristics curves of PV array under different faulty conditions. The proposed technique can be used to identify and classify the five common faults in PV system. These progressive faults (includes open circuit, line-to-line, partial shading, degradation and bridging faults) have been occurred at the basic components of PV cell and at the connection point between PV modules. The characteristic curves under normal and faulty conditions have been compared by utilizing Matlab/Simulink environment. Hence, the suggested method can be utilized to determine the important fact about the health PV array.

Electric Vehicle Routing Problem: Model and Algorithm

Abstract: In this paper, a survey on the related research of electric vehicle routing problem is present. Our review summarized the literature published between 2011 and 2019, in which the variant of electric vehicle routing problem and the algorithms are discussed. In addition, the future trends are analyzed.

Study on the Health Condition Monitoring Method of Hydraulic Pump Based on Convolutional Neural Network

Abstract: Aiming at the different health states of hydraulic pumps, this paper proposes to use the convolutional neural network to classify and identify the time-frequency images of different volumetric efficiency of hydraulic pumps to realize the health status monitoring of hydraulic pumps. After collecting the vibration signal data of the hydraulic pump, the time-frequency map is formed by short-time Fourier transform, wavelet transform, and Wigner-Will distribution, and then the generated time-frequency map is divided into a training set and a test set, in which the former is used to train the convolutional neural network and the latter is used to verify the recognition result. Finally, it is proved that the convolutional neural network can well identify the different health states of the hydraulic pump, and the recognition rate reaches 99%.

Computer Aided Design Simulation of 3D Garden Landscape Based on Virtual Reality

Abstract: At present, landscape design teaching in design art education has received much attention. With the continuous development of social economy, science and technology and information network technology, traditional landscape design has also achieved a "quality" leap. The appropriate use of computer-aided software in landscape design has become an essential skill for landscape designers. Computer-aided software is like a pen in your hand, a tool that shows the designer's ideas and thinking. How to make students master computer-aided design software and use it reasonably in landscape design is a problem that current landscape design teachers need to think about and study. The existing CAD software for garden landscapes has limited expression of 3D models and virtual simulation effects, and can only be used for 2D construction drawing. In order to solve the dynamic landscape characteristics, a three-dimensional landscape design simulation system of virtual reality technology was developed. Object-oriented parametric 3D solid modeling method was used to realize 3D solid modeling, 3D real-time virtual rendering of design scenes, renderings and animations. Production, construction drawing and other functions, and explored the system computing principles, implementation methods and OpenGL extensions and other related technologies. Then the overall function of the system is briefly introduced, and the simulation process in the professional module is discussed in detail. Finally, the effectiveness of the system is verified by an engineering example.

Application Research of Apriori Algorithm Based on Matrix Multiplication in Children's Drug Interaction

Abstract: The Apriori algorithm of association rules requires multiple scans of the transaction database to obtain frequent item sets, which results in the generation of a large number of candidate sets and affects the operation efficiency of the algorithm. In this paper, improved Apriori algorithm based on matrix multiplication is applied to the study of the interaction among children's anti-infective drugs, antipyretic-analgesic and anti-inflammatory drugs, and drugs for digestive system, so as to explore the rule of the combination of three kinds of drugs, and provide theoretical guidance and reference for children's clinical rational drug use.

Mobile Learning System of Ideological and Political Education in Universities Based on Android

Abstract: This paper introduces the research background and key technologies of this project, and chooses Android mobile operating system as the platform of project design and implementation. Through the study and analysis of the theory of mobile learning, it puts forward the demand analysis of the mobile learning system of ideological and political education in universities. Through the research of mobile development technology and mobile learning system, this paper expounds the overall functional framework design of Android client, and explains the design of five functional modules and their sub functions. From the overall point of view, the communication protocol, data storage and server-side design of the system are analyzed. The communication protocol, data storage and plug-in secondary development of the server are mainly described. The test results show that the system can run stably in practical application, which has perfect functions and it is easy to be operated. The scheme meets the friendly interaction of mobile application, and meets the requirements of mobile learning system in practical learning.

Study on Temperature Field of In-Wheel Motor for Electric Vehicle Based on Magnetothermal Coupling Method

Abstract: In order to improve the working performance and service life of electric vehicles, the temperature field of the in-wheel motor is optimized based on the finite element analysis model. The winding copper loss, stator core loss, permanent magnet eddy current loss and stray loss are used as heat sources for the temperature field analysis of the in-wheel motor. Each component of the in-wheel motor is analyzed for transient temperature changes using a magneto-thermal coupling method. The temperature variation of the components of the in-wheel motor with time and the spatial distribution characteristics are studied in detail. Finally, the joint simulation model including the in-wheel motor body and the external drive control circuit is established, which fully considers the transient changes of the motor loss distribution during the working process.

Practice of "Independent-Cooperative-Inquiry"-Based Ideological and Political Education Teaching Model in Aerobics Classes at Colleges and Universities

Abstract: College aerobics classes are closely related to reality and times. They are required to "advance with the times" and match with ideological and political education closely. Ideological and political education provides a new model for aerobics teaching. Based on the summary of aerobics teaching practices, the importance and superiority of the "independent-cooperative-inquiry"-based ideological and political education teaching model in aerobics teaching is clarified in this paper. The main problems in the ideological and political education teaching in aerobics teaching are put forward to further optimize aerobics teaching through ideological and political education and improve the specific teaching effect effectively.

A Sensitive Information Detection Method Based on Network Traffic Restore

Abstract: With increasing importance of protecting sensitive data, it is necessary to monitoring the transmission of sensitive data. Most recent methods detect the network packet based on the analyzing IP, port, size characteristics and payload of the network traffic packets, which can not be used for detecting sensitive document transmitted via network. Hence, this paper propose asensitive information detection method based on network traffic restore solution to fix the problem. This method firstly use TCP reassembling to restore documents from network traffic, and then use TF-IDF model to judge the similarity with those documents marked with sensitive flag, which is very simple and efficient

Driver Gaze Region Estimation Based on Computer Vision

Abstract: Traditional methods of driver's gaze estimation usually need additional equipment to obtain the driver's facial and eye features, which is difficult to apply in real life. In this paper, a gaze region estimation method based on computer vision is proposed, which reduces the hardware requirements of the experiment. In order to achieve this goal, a new eye feature extraction method and an improved random forest algorithm are adopted in this paper. Finally, the driver's line of sight is determined by combining the head and eye features. The driver's head features are obtained by pose from orthography and scaling with iterations (POSIT) algorithm. Then, the accurate positions of the driver's corner of the eye and the pupils are obtained by the from coarse to fine method, and the eye line direction is estimated according to the relative position of the pupil in the eye area. Finally, the improved random forest is used to estimate the driver's gaze region. The experimental results show that the accuracy of this method is 94.12% in the real driver's gaze data set.

A Human-Object Interaction Detection Method Inspired by Human Body Part Information

Abstract: In recent years, with the rapid development of big data and computer hardware, deep learning has regained vitality, and the field of computer vision has made great progress. Human Object Interaction (HOI) Detection is an important subject for image understanding in computer vision, which has not been effectively solved yet. Aiming at the problem that the human pose information is not effectively used in the current HOI detection methods, an HOI detection method that integrates human body part information (HBP) is proposed. By applying human body part information to HOI detection, it makes the best of the relationship between parts and objects that effectively improves detection performance. The experiments and results on Verbs in COCO (V-COCO) dataset show the effectiveness of our method.

An Feature Image Generation Based on Adversarial Generation Network

Abstract: Generative antagonistic network (GAN) was proposed in 2014 to assist in generating realistic visual images, which has become one of the most popular research objects in deep learning in recent years. In the field of image generation, GAN is more effective than the traditional method, but it is difficult to train, unstable network and difficult to convergence. In this paper, GAN is applied in the field of feature image generation, and a new framework is proposed based on the C-SEGAN. By adding additional condition features to generator and discriminator, the similarity of distributed error is learned, and the discriminator is self-encoder, the mean square error loss is added to discriminator, and the generated model generates the specified sample. The model can generate the specified clear image according to the feature conditions. The experimental results show that the method can achieve faster convergence rate and generate better quality and diversity images with a simpler network than other supervised class generation models.

Financial Risk Prediction and Evaluation Model of P2P Network Loan Platform

Abstract: In this paper, the big data crawler technology is used to collect the open data of the third-party website platform of online loan, and the machine learning model is used to study the risk of illegal fund-raising of online loan platform. The traditional machine learning method represented by logical regression and decision tree model and the frontier machine learning model represented by random forest model and Lightgbm model are compared in multiple pre-warning. In the framework of dynamic early warning, the dynamic early warning effect of each model in the whole life cycle of online loan platform is studied. The results show that both the traditional and the frontier machine learning model show excellent early warning effect. The accuracy of the traditional model is slightly lower than the frontier model, but the performance of the decision tree model in the important detection rate index is better than the other models.

Assessment of Car to Car Crash Compatibility Based on MPDB

Abstract: In order to maximize the balance between the vehicle's aggressiveness and crash worthiness of different vehicles during the collision, the car-to-car crash compatibility of the new MPDB test is taken as the research object. A mathematical theoretical model of vehicle-to-vehicle collision was established, showing that the main factors affecting vehicle collision compatibility are: vehicle mass, front-end stiffness, and structural size, and The speed variation of the two vehicles is inversely proportional to the mass, and the energy absorbed by the vehicle is also inversely proportional to the average stiffness of the front-end structure. Combining the statistical analysis of different MPDB test data, this paper explores the correlation between influencing factors and collision compatibility, compares the stiffness and energy differences between MPDB and ODB honeycomb aluminum barriers. And the performance and improvement of three compatibility criteria are also obtained, including honeycomb aluminum homogeneity, trolley occupant load criterion and bottoming out.

Accurate and Fast Wavelength Demodulation for Fbg Reflected Spectrum Using Multilayer Perceptron (Mlp) Neural Network

Abstract: An accurate and fast wavelength demodulation method for fiber Bragg grating (FBG) reflected spectrum based on multilayer perceptron (MLP) neural network is presented. The wavelength demodulation is treated as a supervised regression problem and different MLP models are trained by ideal reflected spectrum of FBG under different spectral resolution. The simulation and experimental results show that the algorithm proposed has higher precision and stronger robustness to spectral resolution compared with correlation and bandwidth middle-point (BMP). Moreover, MLP can still achieve an accuracy below 10 pm even when the spectral resolution is 100 pm, which is about 10 pm higher than the BMP method and about 20 pm higher than the correlation method. Meanwhile, well-trained MLP can achieve µs-level fast demodulation for per reflected spectrum. Therefore, the MLP can be considered as an attractive method to extract wavelength for FBG reflected spectrum, especially at coarse sampling, which can achieve fast demodulation in large-scale distributed FBG sensor networks.

Efficient Selection of Visual Features in Automatic Image Retrieval

Abstract: Feature selection is a process of finding an optimal subset of features from the original features set. It could solve the problem of the dimension disaster caused by high-dimensional features, which seriously affects the efficiency of the content-based image retrieval. This paper presents a method for generating an efficient feature reduction method of visual features with neighborhood rough set. By introducing the upper and lower approximation definition of neighborhood rough set, we calculate the approximation information to measure the relevance of the visual features. When the attributes of visual features are reduced and the rest of features can correctly describe the context of the images, the efficiency of the image retrieval could be improved. Furthermore, we use the selected the efficient visual features to perform the image retrieval. Experiment results show that the proposed algorithm is effective in comparison with the other mentioned methods.

Leakage Flux-Based Method to Identify the Inrush Current in Transformers

Abstract: Inrush current occurrence in the power transformers is an unavoidable event that can happen due to energizing no-load transformer. Although various protection methods based on the terminal current and/or voltage have been proposed for transformer protection, the protection malfunction caused by the inrush current still occurs occasionally. In this paper, By comparing different closing angles of transformers with normal condition and differences in leakage flux near the yoke, a simple and accurate method is presented to identify transformer inrush current conditions. The change of the flux density on the leakage flux sensor can effectively identify the transformer inrush current. This identification method has been initially verified in the simulation in Maxwell.

Research on the Application of Big Data in Book Management

Abstract: There are some problems in the management of book materials under the background of big data. The collection of book materials is incomplete, the degree of information management of book materials is not high, and there are missing and duplicated books. Hidden information security hazards such as commercial theft, information theft, and malicious attacks have severely restricted the development of library information management. However, the number of infrastructure facilities cannot meet the actual needs of digital libraries. A management method of books and materials under the background of big data is put forward, that is, focusing on the innovation of knowledge services to maximize the value of knowledge services. Pay attention to the innovation of management mode and use big data technology to build electronic libraries. Under the background of big data, library managers can build an informatization management mode to increase the overall rate of book management.

An Improved Apriori Algorithm Based on Matrix

Abstract: This paper analyzes the limitations of the Apriori algorithm for mining frequent itemsets with low time and space efficiency, and proposes a frequent itemset mining algorithm for matrix, itemset count and item index list. It only needs to scan the database once. According to the prior nature of frequent itemsets, the size of the data scan is reduced by compressing the matrix, and then the bitwise and operation is performed on the compressed matrix row vectors. The itemset count and prefix index list are used to generate frequent itemsets. No candidate set is generated during the mining process. This algorithm improves the efficiency of the original algorithm.

Research on 3D Landscape Design Simulation Platform under Virtual Reality Architecture

Abstract: Taking the garden design of a city square as an example, a garden roaming display system is developed by combining quest 3D technology. This paper introduces the design process of landscape roaming display system, and studies the modeling of landscape scene, the control of plant and water model, the setting of light and shadow, the setting of camera, etc. We discusses the application of quest 3D technology in landscape design, studies the related technology of landscape interactive display, summarizes the method and process of constructing landscape roaming display system, and provides an effective method and way for the display of landscape design Finally, this article establishes a garden virtual reality simulation display platform system. The experimental results show that the system can achieve better results in general mainstream PC, and it has certain reference value for other roaming demonstration systems.