With the development of economical society, the problem of product piracy security is becoming more and more serious. In order to protect the copyright of brands, based on the image neural style transfer, this paper proposes an automatic generation algorithm of anti-counterfeiting logo with security shading, which increases the difficulty of illegal copying and packaging production. VGG19 deep neural network is used to extract image features and calculate content response loss and style response loss. Based on the original neural style transfer algorithm, the content loss is added, and the generated security shading is fused with the original binary logo image to generate the anti-counterfeiting logo image with higher recognition rate. In this paper, the global loss function is composed of content loss, content response loss and style response loss. The L-BFGS optimization algorithm is used to iteratively reduce the global loss function, and the relationship between the weight adjustment, the number of iterations and the generated anti-counterfeiting logo among the three losses is studied. The secret keeping of shading style image used in this method increases the anti-attack ability of the algorithm. The experimental results show that, compared with the original logo, this method can generate the distinguishable logo content, complex security shading, and has convergence and withstand the attacks.
Recently, facial expression recognition (FER) has become an important topic in computer vision research. With the advance of artificial intelligence, the performance of model about FER has made great progress and improvement. To further enhance the ability of extracting significant features and enhancing the robustness of the model, we present a innovative facial expression recognition framework based on convolutional neural network and attention module. Concretely, we add the L2 norm features in CBAM and re-scale the channel weights. Salient attention block is used to suppress the insignificant feature and enhance the weight of salient features, which improves the performance and robustness of the model. Finally, without using extra training data, IVSA achieves the highest single-model accuracy of 72.44%, which is improved by 1.14% compared with the previous methods. Extensive experiments prove the effectiveness of the model and framework.
Galloping of overhead transmission lines is a kind of instability phenomenon created by the wind excitation and asymmetric ice layer on the conductor. It may lead to the damage of overhead transmission lines. In this paper, the aerodynamic characteristics of four kinds of beam conductors with aerodynamic dampers are studied on the basis of wind tunnel tests. To perform such analysis, the aerodynamic coefficients for four bundle conductors with different aerodynamic dampers under different wind directions and different ice-coat were obtained. The tests showed that the drag coefficient of leeward conductors increased about 10% while the lift and moment coefficients decreased about 46% and 23%. The numbers of negative Den Hartog coefficients are respectively about 22.2% (different attack angles) and 70.0% (different ice angles) less than the case without any aerodynamic damper.
Over the past decade, motion planning in robotics has become not only a popular research field in academia, but also one of the most applicable technologies in various industries. There are two famously known approaches to tackling this problem, namely, the search-based approach and the sampled-based approach, both involving the idea of reposing a motion planning problem as a search problem in the Configuration Space. The main focus of this report is to illustrate how a search-based approach, originally proposed to solve a search problem, can be accommodated for the use of planning the motion of a simulated two-link arm. Moreover, the performance analysis of three different search algorithms, Depth First Search (DFS), Breadth First Search (BFS), and A-Star (A*) for a grid search will be provided to show the conclusion that A* is the best search-based algorithm out of the three to use in practice because it generates the most efficient arm movement with a slight tradeoff in a runtime.
The dielectric spectrum can effectively reflect defects such as the aging and dampness of the cable, but the traditional measurement of the dielectric spectrum of the cable needs to be carried out after the cable insulation is sliced, which is very cumbersome. This paper proposes a fast cable dielectric spectrum measurement method, which can directly measure the dielectric spectrum of the entire cable without slicing. In the paper, the principle of the measurement method was introduced, the thermal aging and water tree cable samples were prepared separately, and their dielectric spectra were measured using the proposed method. The test results show that the measurement result of the method proposed in the paper is greater than that of the traditional dielectric spectrum measurement method. It can effectively detect and diagnose thermally aging cables, but the detection effect of water tree defect cables is not obvious.
Aiming at the low efficiency of image target detection in cloud computing mode, a target detection system suitable for edge devices is designed. First, the system selects Faster R-CNN in the deep learning algorithm as the target detection and recognition model, and trims the network feature extraction layer through the residual module. Second, a proposal region extraction sub-network with adjustable anchor boxes is used to obtain proposal regions more quickly by setting a convolutional sliding window of reasonable size. Finally, a complete target detection system is built using hardware such as Raspberry Pi development board and Intel neural computing stick. The experimental results on the KITTI dataset show that the system achieves good detection results, and achieves a faster recognition speed without reducing the target detection accuracy, which can meet the real-time requirements of offline work.
The traditional construction mode of 35kV substation has many problems that need to be solved urgently, such as long construction period, large area and large operation and maintenance workload. In order to solve the above problems, this paper proposes an integrated method of intelligent assembly of bench-type transformers. This paper adopts key technologies such as prefabricated buildings, modular electrical secondary equipment, prefabricated optical cables and cables. The substation adopts a combination of factory prefabrication and on-site installation in the construction. This makes the structure of the substation more compact. At the same time, this paper proposes a 3D visualization component assembly technology for smart substation simulation. The visualization component includes the three-dimensional model of the equipment in the smart substation, and the simulation realization of the equipment's logic function and the IEC61850 communication module. The system can not only quickly complete the three-dimensional design of a smart substation by assembling visual components, but also realize debugging and simulation of the entire substation by connecting and configuring the underlying modules of the visual components. Experiments show that this technology can improve the development efficiency of a new generation of smart substation 3D simulation system.
In the era of big data, digital recognition technology is particularly important, and with the continuous upgrading of relevant information technology, beyond recognition technology has been widely developed in various fields. Through computer information technology as a way of information transmission.it is the basis of digital image processing technology as well as related recognition technology. In the "metaverse" environment, a large number of outstanding paintings are made into digital images and used for "big data" recognition. In order to find such artwork images better and faster, a more scientific and effective recognition method needs to be explored. In this paper, we compare the advantages and disadvantages of feature extraction and classification design with existing methods, and improve and optimize the feature extraction and classification recognition algorithm after completing artwork pre-processing. Finally, the performance of the method is verified by experimental analysis.
To actively respond to the strategic goal of carbon peaking and carbon neutralization proposed by Chinese government, the reform of heat supply is an important research hotspot. Intelligence control technology is used to propose the adaptive distributed heat suppling strategy. Firstly, the data monitoring platform with distributed control points performs as the core. The data resources are sampled and analyzed to realize the visual modeling of distributed databases. Secondly, the platform of distributed databases is established to model the underlying data as a whole for realizing the data- driven model. The preprocessing, filtering, cleaning and mining processes of data resource establishes a multi-level model for analysis of running data. Lastly, the application of the heat suppling strategy not only overcomes the overall variation in practice, but also optimizes the data structure and realizes comprehensive data analysis for different objects. It provides technical support for the green, low-carbon and energy-saving running of the enterprise for the goal of "carbon peaking & carbon neutrality".
The automatic driving robots can replace vehicle drivers in traditional car tests and ADAS tests, and can also realize automatic driving of vehicles. Based on the kinematics and dynamics model, the trajectory planning of the vehicle can be more convenient to control the vehicle speed and path tracking, and at the same time can control the car to do some high-risk tests. This program first establishes the vehicle kinematics and dynamics model, based on the MPC strategy, and integrates the lateral and the longitudinal control, and finally drives the robot to realize the control of the vehicle. The self driving robot uses Visual Studio to write the upper computer software, which has achieved excellent human-computer interaction effect, and verified this statement through vehicle testing. The experimental results show that the lateral control strategy based on model predictive control can well realize the real-time control of vehicle steering by the robot, and the longitudinal precise control of the vehicle by the robot can also be realized based on the kinematics and dynamics models.
The large diameter percussive reverse circulation drill is widely used in large buildings, bridges and other projects in the drilling process. The structure fracture and weld cracking of the drill bit often happen because of the long suffering the effect of cyclic loading cycles. It is the primary cause for the low service life of the drill bit. This paper uses ANSYS nCode DesignLife fatigue analysis software to evaluate and predict the fatigue life of the drill bit under normal working and eccentric load conditions. It Obtained the fatigue life of the drill bit and the dangerous point of fatigue failure. The structure of the original drill is optimized, which provides theoretical guidance for the further improvement of the large-diameter impact reverse circulation drill.
In order to realize on-line monitoring of metering devices and real-time collection of important information of users, this paper proposes the overall structure model of the design and implementation of a remote power data acquisition system for a power supply company, and proposes various items that the system should meet according to actual needs. The technical structure, business function structure and network structure of the power data remote acquisition system are designed, and the main problems existing in the installation, commissioning and system operation and maintenance of the system are studied, especially the main index requirements and troubleshooting of on-site operation. The design integrates various advanced communication technologies such as GPRS/CDMA/SMS, low-voltage power line carrier, etc. with large-scale database, embedded computer and other technologies. Based on B/S and C/S modes, the main station, data acquisition layer, the design of the three-layer structure of the measurement monitoring layer and the complete realization of the system functions. A set of power user electricity information collection system integrating basic application, advanced application, operation management, statistical query and system management has been successfully established, which fully meets the needs of electric power enterprises for electricity consumption information collection and management of power users.
The insulator that plays an insulating role at one end of the high-voltage line is exposed to the air environment, and long-term pollution accumulation in a humid environment will cause flashover between the insulators, which will affect the power supply line and require greater cost to detect and repair. At present, the contact measurement method is used to detect the pollution on insulator surface by cleaning, and the non-contact measurement method uses the idea of characteristic band imaging, which can guide the regular cleaning of insulator surface. In this paper, a method is proposed to use layering ideas and machine learning algorithms on hyperspectral image of a sampling cloth wiped clean from the insulator surface to achieve insulator pollution detection. Hyperspectral data is a three-dimensional data cube with a large amount of data. Firstly, the idea of stratification is used to identify the interval where the ratio of non-soluble deposit density (NSDD) and equivalent salt deposit density (ESDD) of insulators is located, and then XGBoost machine learning algorithm is used to establish the regression detection model of salt density and grey density in the corresponding interval. The method has been proven to have high accuracy and feasibility, and can guide the regular cleaning of insulator surface pollution in practical applications.
High-voltage circuit breakers are key devices in high-voltage power systems, controlling both the on-off of specific lines and the reliable disconnecting of fault areas in the event of a fault to ensure the safe operation of other areas. The mechanical movement of the repulsive device and the force analysis are particularly important in the key structural model of the high-speed opening and breaking of the circuit breaker. In the process of high-speed stamping and resetting mechanism, the key components in the valve body, such as the spool, are subjected to the pressure of the hydraulic fluid with constantly changing pressure. The value of the pressure will not always be constant. The impact force on the mechanism is fast and instantaneous, while the fluid force is continuous, so the analysis of the fluid force becomes critical in the inertia phase and buffer phase of the repulsive mechanism movement. To address this issue, in order to obtain simulation data results closer to the real working conditions, we conduct a study of fluid-structure interaction simulation for structures in repulsive devices.
This paper proposes research on the prediction and health management system (PHM) of aircraft conditional maintenance based on massive data mining. This paper reviews the development history and development status of PHM technology in foreign aircraft engines, fixed-wing aircraft, helicopters, aerospace vehicles, ships, vehicles and rail transit equipment, and then according to the status quo, from the perspectives of system structure, service environment, data sources and storage methods, etc. The application requirements of new generation aircraft for massive data mining are analyzed. Thirdly, based on the data mining work of prediction and health management, this paper proposes a device life prediction method based on Gaussian Hidden Markov Model. The method is divided into an offline stage and an online stage. The offline stage extracts sensor data features and trains model parameters. The second stage then utilizes the model to assess the current state of device health. Finally, this paper proposes an embedded heterogeneous design aircraft PHM massive data mining platform. And designed the overall framework and hardware and software structure of the platform. Finally, this paper takes the aircraft bearing as an example to verify the effectiveness of the algorithm and data platform proposed in this paper. The experimental simulation shows that the data mining efficiency of the system is high, and the accuracy of the equipment life prediction method is as high as 95%. The platform has great military application significance for promoting the integration and engineering realization of aircraft PHM.
This paper studies KNN algorithm and analyzes the factors that affect the accuracy of image classification. Then the algorithm is improved by optimizing the selection strategy of K value in traditional KNN model using the characteristics of genetic algorithm, which is survival of the fittest and survival of the fittest. The Fasion-MNIST dataset is selected to extract the gray pixel value of the image. KNN algorithm model and GA-KNN algorithm model have been trained, and the classification accuracy of the latter is improved by nearly 10%, making the image classification effect better.
With the expansion of the application field of intelligent robot, people expect intelligent robot to serve human beings in more fields and complete more complex work instead of human beings. However, the environment of intelligent robot is often unknown and difficult to predict. It is becoming more and more difficult to analyze and design the behavior of intelligent robot manually. Combined with the actual situation, this paper analyzes the application of robot technology in the competition, and designs a new design scheme of Wushu challenge arena robot, which carries out innovative design from the aspects of mechanical structure, perception system, control system, strategy response, motion distribution and so on. The results show that compared with the traditional scheme, the structure of this scheme is reasonable and stable, and the motion and computing performance are more outstanding.
In this paper, a fatigue driving detection system is designed. Firstly, the real-time collected facial images of vehicle drivers are pre-processed, and the AdaBoost algorithm classifier is used to locate the faces and eyes. Then the edge detection and contour extraction are performed on the obtained eye images to get the degree of eye opening and closing. Next, the proportion of the eye-closing time to the total observation time is calculated in combination with the perclos fatigue driving standard to determine whether fatigue is achieved. When fatigue is achieved, the system will give an early warning to wake up the driver who is sleepy. The simulation results show that this method has strong real-time performance, high detection accuracy and strong practicability.
Under the background of information age, image processing technology is a commonly used technology in all walks of life. In order to improve the image correction effect, it can be processed based on computer vision algorithm. This paper introduces the image categories and technical characteristics of computer vision algorithm and image processing technology, discusses the image processing technology based on computer vision algorithm, including computer vision display system and image distortion correction two parts, the effect is good, has high application value. In the past, due to the lack of technological development, objects in the traditional two-dimensional environment can only display the side projection, relatively single, in order to adapt to social progress, technology will continue to develop, so three-dimensional three-dimensional picture emerged at the historic moment, a new display technology appeared. In order to obtain the ideal visual effect, technicians need to implement image processing technology on the basis of computer vision algorithm, express the actual coordinates of objects in three-dimensional space with three-dimensional voxels, and correct the distorted image caused by projection. Compared with the traditional BP neural network, the image processing technology based on computer vision algorithm obviously has more advantages and higher accuracy.
With the advent of Industry 4.0 and the Industrial Internet, the Internet of Things (IoT) development of applications is booming, the mining machinery working environment is extremely harsh, therefore, the requirements of the performance, quality, durability, reliability of its gearbox is pretty high, to meet these requirements, real-time monitoring is becoming a demanded task. For this purpose, a gearbox monitoring system based on edge computing is established. In the paper at hand, a novel Jacobi-type data parallel processing method is proposed, with which, the efficiency and life of the gearbox are calculated through the edge service APP. Traditional methods by solely utilizing cloud computing cannot effectively accomplish this task. Using cloud-edge collaboration technology, the Web application in scenarios such as intelligent mining is designed, which can grasp the operating status of equipment in the entire mining area, unify scheduling and orchestration of computing resources, update the monitoring model on edge computing nodes, and process and generate effective data of machinery and equipment in real-time at the edge computing device. It reduces the operation and maintenance cost, solves the problem of monitoring data congestion caused by insufficient data bandwidth, and ensures a stable and safe operation of mining machinery.
