Abstract:In order to realize the accurate measurement and tracking of UAV LOS stabilization system, a LOS stabilization controller based on classical PID and fuzzy control is designed. There are a lot of uncertainties in the process of PID parameter tuning in practical engineering. In order to realize the on-line tuning of PID parameters, this paper combines fuzzy control algorithm with classical PID control, and constructs a parameter self-tuning fuzzy PID controller, which realizes the correction of PID controller. In MATLAB Fuzzy Toolbox and Simulink, comparing PID with parameter self-tuning fuzzy PID, the step response curves of parameter self-tuning fuzzy PID controller under no disturbance and 10 Hz sinusoidal disturbance show that fuzzy PID has shorter response time and smaller isolation degree than fuzzy control and PID control, and the fuzzy PID error occurs in the sinusoidal response curve of the system with input from 1~10 Hz. The difference is the smallest and the control effect is the best. Thus, the self tuning PID fuzzy controller has good robustness and control performance in LOS stabilization.

Abstract:In order to meet the requirements of real-time monitoring aboutstates of battery group, optimizing control strategy, prolonging battery life and ensuring the safety of use, this paper proposed and implemented a battery pack cloud management system using optimized residual power (SoC) estimation.Based on the traditional SoC estimated strategy (charge accumulating method and open circuit voltage), the extended calman filter algorithm wascombined to improve the accuracy of battery pack’s remaining power estimation, and the mathematical model between the battery voltage and the SoC was established.At the same time, the cloud-server system was added to realize the real-time detection of the battery group. In addition, a battery pack management device with a capacity of 22.2 V was designed and implemented. The experimental results show that the SoC estimation accuracy of the battery management deviceis high, the measurement error is less than 1%, the system stability is good, and the requirements for real-time detection of the battery state are met.

Abstract:There are only cracks and depressions characters on the road surface can be detected by the traditional road detection methods. In order to detect the internal conditions of the road, the self-powered wireless sensor network is used to collect the internal temperature and humidity data of the road. The wireless sensor module is protected by epoxy resin and buried in the road, and the road load energy is collected by the self-powered module and used to power the wireless sensor node. The sensor begins to be used to collect data that be transmitted to an external data processing center via the wireless sensor network while enough energy had been collected. The experimental results show that after four months of data collection, the wireless sensing system embedded in the asphalt road can be applied to transmit data for a long time, therefore, the current road structure health status also can be analyzed through the collected road temperature and humidity data.

Abstract:In order to realize the intelligent control of the greenhouse environment system better, aiming at the problems of non-linear, strong coupling, large lag and strong time-varying in greenhouse environment system, this paper proposes and designs a BP neural network PID controller based on genetic-particle swarm optimization on the basis of analyzing BP neural network technology.Combining the strong global search ability of genetic algorithm and the strong local search ability and fast convergence speed of particle swarm optimization algorithm, the controller optimizes the weights of neural network and effectively controls the greenhouse environment system.Finally, a comparative study of conventional and improved BP neural network PID controllers is carried out.The simulation results show that the improved BP neural network PID control has better stability and robustness.

Abstract:Since the beginning of the 21st century, the concept of using robots to replace manual work has been widely recognized all over the world. The development level of robotics technology has become a representation of a country’s high-tech industry, and intelligent security robots as a member of many robots has been widely concerned. In this paper, a powerful and stable control system is designed for the intelligent security robot. Firstly, the overall structure of the intelligent security robot control system is designed by using modular design ideas. Secondly, the main controller of the robot control system is designed and developed, and the motion control subsystem of the robot is also designed. Besides, sensor subsystem and camera control subsystem are researched and designed. Finally, the control system is tested by the robot platform, and the experimental results are analyzed and summarized. It can be foreseen that the security robot is of great significance in reducing the risk of manual operation and the cost of enterprises.

Abstract:Conventional mechanical angle measurement sensors are utilized in order to solve the problems caused by increasing system size, weight, cost and installing sensors that lead in errors and unreliability in frame servo systems. Based on TMS320F28335 as main controller and A3PE3000 as slave controller, we have designed the hardware circuit for electrical signal acquisition and operation. Meanwhile, a sliding mode observer is proposed for angle estimation to replace mechanical sensors. In this method, the relevant electrical signals in the motor windings are measured by hardware circuits. The rotor position and velocity estimation are implemented by establishing a sliding mode observer and a phase-locked loop algorithm. The simulation model is built in Simulink. The simulation results show that the values of position and velocity estimated by the sliding mode observer are convergent to real value after the system runs for 0.2 s.

Abstract:In order to solve the problem of vertical interconnection of inter-layer signals in three-dimensional integrated circuits, based on the excellent characteristics of thick film substrates, an ultra-wideband coaxial to stripline to microstrip transition structure is designed, which can work up to 80 GHz. This vertical structure uses LTCC as the substrate and spin-coats the BCB film for system-in-package. The structure utilizes a “water droplets” matching method and is embedded in the air cavity to suppress parasitic capacitance, which effectively improves the RF transmission performance of the interconnect structure. The simulation results show that the back-to-back (coaxial-stripline-microstrip-stripline-coaxial) structure has a return loss of less than -17 dB in the frequency range of 0~80 GHz, and the insertion loss is better than -0.5 dB, and VSWR are lower than 1.35, which has good ultra-wideband transmission characteristics.

Abstract:A diversified project for electrical fire is proposed in this paper, in order to effectively monitor the electrical fire of low voltage cables in coal mine wellhead. Firstly, it is relay protection, to prevent fire accident caused by cable heating because of electrical fault. Secondly, it is temperature fire detection alarm, smoke fire detection alarm or infrared fire detection alarm, to detect initial fire caused by external high temperature, and to be the backup of relay protection. Thirdly, it is video monitoring, to provide basis of fire determination for the person on duty. To combine the three, can realize the effective monitoring of electrical fire. And it also gives practical engineering applications at last.

Abstract:In order to solve the problem that the stability of the power grid will be affected by the lack of inertia and damping when the double fed wind turbine is connected to the outlet inverter, this paper first describes the whole control strategy of the doubly fed fan based on the virtual synchronous control. Then the control strategies of the rotor-side inverter and the grid-side inverter in the overall control strategy are introduced respectively, and the double-fed fan unit model with grid-side VSG control is built. Finally, the simulation analysis of the model is carried out through MATLAB/Simulink for the fault condition and different running wind speed of the power grid, and the simulation results of the original double fed fan model are compared and analyzed. It is verified that the double-fed fan with grid-side VSG controlhasbetter regulationcharacteristicsthan thedouble-fed fan.

Abstract:Research and application of key technologies for intelligent operation and maintenance of automatic calibration lines

Abstract:In view of the fixed window function of S-transformation and the problem of detection accuracy when detecting the start and end moments of disturbance signal, amplitude variation, and phase change, this paper introduces two tuning factors in S-transformation to obtain an improved one. S transform, and the inverse transform of the improved S transform to get the TT transform, the improved S transform and the TT transform are applied to the three sources of the disturbance that mainly cause the voltage sag, and on the MATLAB to the three a kind of temporary disturbance source signal was detected and analyzed. The MATLAB simulation results show that the proposed method has better effect and higher detection precision than the S-transform method in detecting voltage sags, which lays the foundation for better extraction of the characteristics of sags and identification of temporary sources.

Abstract:In view of the insufficiency of background difference method and interframe difference method in moving targets, the real-time and accuracy of moving object detection are improved. In this paper, a moving object detection method combining background difference and inter-frame difference is proposed. This method simulates the detection method of human eye on moving target, which is divided into two parts: Global perception and precise perception. First, the image is divided into multiple regions and the background model is established using fewer pixels. The background model is used to determine the region where the moving object is located, and the images of other regions are stored as the background image. The current image of the changed region is then used to perform a differential operation with the stored background image of the region to obtain a clear moving target. This method uses less pixels to construct the background model, which reduces the amount of calculation of the background model establishment and update, and improves the speed of the operation. By separating the stored background image from the current image, a complete moving target can be obtained, avoiding the appearance of “holes”.

Abstract:Although great progress has been made in image style transfer based on deep learning, these methods were not took into account the distortion of lines in the generated image. Therefore, new method combining histogram loss with transformed Gramian matrix is proposed. The use of histogram loss in image style transfer is not only enhanced the image, but also made the generated image more stable. Transformed Gramian matrix is similar to Gram matrix, but the former is extracted more complete texture information, and also took into account the spatial arrangement of an image information. The experimental results show that combination of two methods can not only make the generated image without line distortion, but also reduce the number of iterations in image generation.

Abstract:A target detection method based on improved Gabor filtering and regional growth is proposed for the influence of the relative motion of the background, the similarity of the object and the change of the external light. First it will do subtraction between two adjacent frames of the input video image, and then the L component of the Lab model in the difference image is subjected to Gabor filter processing,then we need to choose the right parameters according to different scenarios to extract the salient region of the target. In order to consider the operating speed and the significant features of the target,we need to select a pixel with a gray value as the seed point,and for the purpose of obtaining a more accurate target,we need to perform a region growing operation at a point within a certain range of gray value difference.Next,we need use corrosion and swelling to remove interference. Finally,to achieve the purpose of testing, the eight directions of filtering should be merged together, and the circumscribed rectangle is used to mark the target area that meets the requirements. Through parameter adjustment and experimental verification, the target detection rate of this method is improved to 90.89% compared with the low accuracy of the traditional detection algorithm, and it can have good robustness under the interference of external factors such as illumination and background.

Abstract:In order to improve the ability of data retrieval and mining in large-scale information management system, a data mining technology of large scale information management system is proposed based on semantic association feature extraction. The cloud storage model is constructed to design the big data distributed storage in large information management system, and the optimized structure of the information management system is reorganized with the feature recombination method of big data information flow. The semantic association dimension feature quantity of the information management distribution data is extracted from the reorganized information management system topology, and the integrated scheduling and data mining of the information management system is carried out using the semantic association feature quantity as the training sample set. The fuzzy C-means algorithm is used for adaptive fusion and clustering of semantic association features of distributed data in large-scale information management system, and the feature compressor is used to reduce the dimension of storage space of large information management system. Improve the ability of target data mining and adaptive scheduling of information management system. The simulation results show that the method is accurate and semantic association clustering is strong, which improves the retrieval and scheduling ability of the target data in the information management system.

Abstract:This paper proposed an improved gradient smoothing boundary algorithm for the ringing effect generated by coded aperture imaging in image restoration. As the first step, the blurred image is filtered by means, and the gradient smoothing image is calculated. Then the image is extended by reflection interpolation, and the preprocessed image is obtained by the optimal window processing. Finally, the original image is restored by the improved Wiener filter algorithm. Experimental results suggest that the improved algorithm can effectively suppress the ringing and retain the details of the image. Compared with the gradient smoothing boundary algorithm, the structural similarity of image restoration is improved by 27% and the image quality index is increased by 19%.

Abstract:Noise type recognition for noise images is one of the key techniques for targeted denoising of noise images. Support vector machine (SVM) is a classification method based on statistical learning theory applicable to finite sample cases, and its classification ability depends largely on its related parameters.In this paper anew method is proposed to optimize the parameters of SVM. Grey wolf optimization (GWO) algorithm is used to optimize the parameter of SVM for obtain the optimal classification model，meanwhile the proposed method is applied to the noise type recognition experiment of noise images. The images withnoise interference are formed by three types of noises such as Gauss noise, Salt-and-Peppernoise and speckle noise. 90 sample data are taken as training samples and the remaining 60 sample data are taken as the testing samples. The Zernike moments and wavelet high-frequency non-significant coefficient subband energy ratio are selected as the eigenvalue. The GWO-SVM classifier is used to classify noise images. The experimental results show that the GWO-SVM method has better classification accuracy than the traditional SVM classifier.

Abstract:Edge detection was an important step in image processing, in order to effectively suppressed noise and preserved edge information, analyzed the defects of Gauss filter was not adaptive, according to the characteristics of discontinuous noise gray value based on improved traditional mean filter, proposed a weighted mean filter coefficients adaptively, and through combined with LoG operator, a dual filter edge detection algorithm based on LoG operator was obtained. Compared with LoG operator and Canny operator, this algorithm has better detection accuracy while keeping the detection speed unchanged.

Abstract:In recent years, great progress has been made in controlling the categories or attributes of generated images by adding condition tags to the generation adversarial networks. However, the accuracy of the category or attribute of generated image needs to be improved. In order to solve this problem,we add reinforcement learning to the generator of generation adversarial networks, which guides the current classification by previous. In addition, the attention mechanism is used which makes a global sensory field to the images with only a small amount of computational loss. We combines multi-attribute star generation adversarial networks with self-attention generation adversarial networks which improves the quality of generated. maximum mean discrepancy reaches to 0.036 93 and the 1-nearest neighbor classifier has a batter effect by reinforced-combined generative adversarial networks, which can generate the art images that certain attributes are assigned automatically and accurately. The generated images can also be used to address the lack of data.

Abstract:For the weakness of maximum margin criterion and margin fisher analysis in the process of human face feature extraction, this paper presents margin discriminant projection algorithm. We define within-class scatter matrix using class samples’ mean and it’s marginal samples of same class, and define the between-class scatter matrix using class samples’ mean and it’s marginal samples of other classes. At the same time, the maximum margin criterionis used to solve singularity of within-class scatter matrix. Compared with the classical maximum margin criterion and margin fisher analysis algorithm, margin discriminant projection can consider the globaland local structure of samples at the same time, avoid small sample problem. The experiments on the face datasets show that the margin discriminant projection is a kind of effective feature extraction algorithm and has improved face recognition accuracy.

Abstract:The scattering of defects on the surface of the mirror directly affects the performance and accuracy of the optical testing system. The scattering of the surface defect of the mirror can be detected by the laser scattering microscope. In order to ensure the detection efficiency and precision of the whole system, the microlens lens, the size of the spot and the power spectrum response of CCD are designed in the design of the hardware system. The software of microscope imaging test system can use step motor to scan the mirror surface, control the image acquisition card to collect the image and save the image. It focuses on how to accurately and accurately collect the sub aperture images, and lay a foundation for subsequent image mosaic. By analyzing the accuracy of the whole test system, it is able to distinguish 5~10 μm defects by ×20 magnification.

Abstract:Pedestrian feature extraction is one of the key steps in pedestrian recognition. The traditional method of pedestrian recognition is to extract feature descriptors(such as HOG,LBP) from each color channel (R, G, B channels), Finally merge into a feature vector. it is difficult to take account of the correlation information between different color channels. In this paper, we use a holistic approach to extract local feature descriptors from color images, which is called quaternionic local ranking binary pattern local descriptor (QLRBP). Unlike traditional methods, this method extracts LBP features from the quaternionic representation space instead of the three color channels. First, Encoding a color pixel using a quaternion to get the quaternionic representation (QR) of the color image which collected from a vehicle mounted camera. Then, Applying a Clifford translation to QR of the color image. Finally, Performing a local binary codingon the phase of the transformed result to generate local descriptors of the color image. QLRBP is able to handle all color channels directly in the quaternionic domain and include their relations simultaneously. In the method of pedestrian recognition, the positive and negative samples are collected first. The QLRBP features are extracted from all the samples, and the K-nearest neighbor algorithm is used to train the classifier. The method is tested on the INRIA pedestrian database and shows that it is better than other features, such as HOG features and traditional LBP features. Performance approach to the current advanced method of pedestrian recognition.

Abstract:Compact descriptor for visual search (CDVS) provides a standardized bitstream syntax for mobile image retrieval and matching applications. Although the standard CDVS algorithm has significant advances in the images with the good lighting condition, this approach often fails on dealing with the low-light images. A CDVS matching algorithm based on image enhancement is proposed. Firstly, a histogram equalization method is used for improving the quality of the pictures on low-light conditions to increase the matching number of the key points. Then, a same image enhancement method, which is mainly according to the homomorphic filtering, is adopt as the preprocessing for enhancing the contrast of the image with suppressing the low frequency signal and highlighting the high frequency signal to better extract the descriptors. The experimental results are compared with the image matching results under dim conditions and the image matching results after processing, and the effectiveness of the algorithm is verified.

Abstract:With the increasing status of oil pipelines in the national economy and the inevitable aging and stealing of oil, pipeline leakage monitoring has increasingly become an important guarantee for pipeline operation safety. In order to timely and accurately report the location of the leaking point, a virtual instrument was built to monitor and detect it. Through the improvement of the pressure gradient method, the corresponding software system model was designed. Based on the LABVIEW platform, the model can accurately monitor pipeline leakage, and can accurately locate the leaking points by comparing within and among the groups, thus to improve the accuracy of pressure gradient method. By this improvement, the frequency of false positives is reduced, and the function of real-time monitoring and positioning of pipeline leakage is realized.

Abstract:In recent years, more and more multi-circuit cable lines of different voltage levels are laid underground of the cities, multi-circuit can cause the problems of circulation loss on cable metal sheath become severe, the different voltage levels of the cable lines are mixed laid, which have a greater impact on circulating current. Based on the principle of electromagnetic induction, calculating the induced voltage of 10, 110 kV single core cables in this paper, both of sheaths with applied cross-bonding connection. Circulating current matrix equation is figured out by the impedance model, and finally get the result of circulating current on the metal sheath of different voltage cable lines and the interdependent of each other. The results show that mixed-laying will lead the 10,110 kV line circulating current increase and decrease; 10, 110 kV line’s circulating current will increase with the increasing of space between different phases, but interaction between them is different,10 kV line circulating current increased by 48.99% when 110 kV line phase spacing increased 200 mm,10 kV line phase spacing increased 200 mm, the 110 kV line circulating current almost did not change; the arrange of three segments length of cable on cross-bonding unit will affect their own circulating current, but does not impact the circuit which next to; the varies of load current influence circulating current greatly; The greater the vertical distance between the 10,110 kV line, the smaller the circulating current is; the circulating current is the largest as the phase sequence is BAC-CAB(10~110 kV), the smallest circulating current appears when the phase sequence is CBA-ACB(10~110 kV). The results above will have good reference for cable route planning and design.

Abstract:The temperature and humidity of soil are the key parameters affecting the growth and development of crops, soil temperature affects the growth and development of plants and soil.The formation of various biochemical processes in soil, such as the biochemical and inanimate chemical processes caused by microbial activities, is affected by soil temperature.Soil moisture is also a basic condition for crop growth and development and an important parameter for crop yield prediction. Therefore, we should get information on soil temperature and humidity to formulate manual intervention measures.Application is an important guarantee for stable production. In this paper, AT89S52 microcontroller as the main control chip, combined with temperature and humidity sensors and computers, built a set of farmland soil temperature and humidity monitoring hardware platform;LabVIEW, a software development platform of National Instrument Corporation of America, designs and compiles the host computer software based on the state machine program framework, which realizes friendly users.Interactive interface, complete the real-time measurement of soil temperature and humidity, display and record functions, batch data storage and management, using statistics.Data analysis, data query results draw statistical histogram, generate reports. By using the function of the front panel network release, the local area of the real-time monitoring system is realized network remote control. Tests show that the system works reliably, has high measurement accuracy, simple structure, stable operation, easy to debug and expand, and can accurately and accurately measure soil temperature and humidity.