Abstract:As one of the biophysical parameters in terrestrial ecosystems, land cover is a basic variable that supports various scientific studies such as urban dynamic monitoring. For land cover products based on the IGBP system, this paper proposes an upscaling method based on semantic proximity and clustering rule sets. And on this basis, in view of the lack of real reference data in the current upscaling evaluation method, an accuracy evaluation index ICS based on Structural Similarity is designed. The experiments based on the MCD12C1 data of MODIS were carried out, and the results of different upscaling methods were comprehensively compared, analyzed and discussed using the proportion method and ICS respectively. The designed semantic proximity method and rule set method reached 84.1% and 84.8% in weighted ICS, respectively, and the feasibility and effectiveness of the proposed method were verified.

Abstract:It is aimed at the phenomenon that the parallel current sharing effect of the buck(Buck) converter in the chaotic state is not good, and the steady-state and dynamic characteristics of the current sharing are poor. Firstly, the segmented smooth switching model and discrete model of the voltage-controlled Buck converter are established, and the path to chaos is analyzed. Then the drive and response systems with different initial values are established. The designed terminal sliding mode control rate is applied to the response system to realize the chaotic synchronization control of the parallel system. Based on the chaotic synchronization characteristics, the parallel current sharing control of the voltage-controlled Buck converter can be realized. Finally, the parallel system model after applying control is built through the Matlab/Simulink simulation platform for experimental simulation. The results show that the current sharing error of the parallel system is close to 0, and when the input voltage changes abruptly, the current sharing error does not change, indicating that after applying terminal sliding mode control, the Buck converter parallel current sharing error is small and has a good steady state and dynamic characteristics.

Abstract:In view of the shortcomings of traditional waveguide-slot antennas with narrow bandwidth, many layers of cavities and high profile, a new low-profile and wide-band waveguide-slot antenna is designed. This design improves both the radiating element and the feeding network of the traditional scheme, increases the gap spacing, as well as achieves the effect of good broadband matching by reasonably increasing the allocation block. The increase of the slot spacing enables the same-layer design of the feeding network to be realized, thereby reducing the number of layers of the antenna and reducing the antenna section. In this paper, the HFSS simulation optimization of the new flat-panel antenna is carried out. The simulation results show that the relative bandwidth of the new flat-panel slot antenna can reach 15%, and there are only 2 layers, with the advantages of wide frequency band, less layers and low profile. The antenna can be used as the main solution for the future flat-panel slot antenna.

Abstract:In order to solve the problem that the jitter value of the high-frequency digital signal inside the chip is difficult to measure accurately, this paper proposes an expander circuit structure that can reduce the frequency of the high-frequency digital square wave signal. The expander samples and outputs the edge of the high-frequency digital signal, and expands the input high-frequency digital signal into a low-frequency square wave signal with a preset period in real time while completely preserving the signal jitter. Transistor-level simulation experiments and MSI (Medium-Scale Integration, medium-scale integrated circuit) board-level verification show that this method can expand the time interval between adjacent edges of the signal, while preserving the jitter characteristics of the original signal, and can be used to measure frequencies up to several The jitter of high-frequency digital signals in gigahertz and the measurement accuracy is very high (error less than 0.7%). The expander has a simple structure and can be integrated inside the chip to quickly and accurately measure the jitter of the high-frequency digital signal on the chip.

Abstract:In order to achieve the quality control of the detection equipment-level data of the regional automatic weather station, and to improve the stability, authenticity and reliability of the weather detection data from the observation source, this paper is based on the existing technical architecture of the regional automatic weather station, using 3 temperature sensors, 3 rainfall sensors, 2 A wind sensor and an intelligent control unit transform the existing regional automatic weather station, which mainly realizes the multi-sensor redundant observation of the non-linear variables (temperature, rainfall, wind direction, wind speed) of the meteorological observations. By setting the front-end algorithm of the collector, Perform threshold cross-validation on multiple detection data of the same element at the same time to obtain the optimal detection data.The test results prove that the multi-sensor redundant observation technology can effectively improve the uncertainty and instability of single-sensor measurement results, and it is of great significance to improve the availability and accuracy of detection data from regional automatic weather stations.

Abstract:Considering the shortcomings of large volume and high power consumption of the current multi-channel fiber Fabry -Perot pressure sensor demodulation system, this paper proposes a method to realize four channel demodulation through rapid and continuous switching of MEMS optical switch, selects a low-power CMOS image sensor s11639, and constructs a hardware demodulation circuit with FPGA as the core, The multi-channel switching logic constraints and the acquisition of interference spectrum are completed by writing s11639 and ad9826 drivers. Finally, the cavity length of the sensor connected to each channel is displayed in real time in the host computer through data processing and transmission. The calibration coefficient is determined by the standard demodulation instrument, and the cavity length of the four pressure sensors connected to the access is actually measured. There is no jumping point in the process of cavity length measurement, which shows that the four channel demodulation system designed in this paper has good calibration stability and small error. The nonlinearity of each channel is calculated according to the measured data. The maximum nonlinearity of the four channels is 0.93%, and the demodulation speed of each channel can also reach 10Hz.

Abstract:The heading angle error is one of the main errors in pedestrian navigation. A common solution is to use the MIMU in combination with magnetometers, but the magnetometer has a large error, so it needs to be calibrated before use. To solve this problem, this paper proposes a pedestrian navigation algorithm based on magnetometer online calibration. Before walking, the ellipsoid fitting through spatial rotation is used to complete the preliminary calibration, and then the UKF is used to estimate and eliminate the time-varying error in real time during the journey. Correct the pedestrian heading with the output of the calibrated magnetometer. Experiments show that the pedestrian navigation algorithm based on online magnetometer calibration has reduced the closed-loop error by 6.17% and 2.8%, respectively, compared with the pedestrian navigation algorithm where the magnetometer is not calibrated and the magnetometer is only initially calibrated. The results show that the algorithm proposed in this paper effectively calibrates the magnetometer, suppresses the divergence of the heading angle in pedestrian navigation, and improves the navigation accuracy.

Abstract:Selection and parameter design of MIMO radar waveform directly determine the key radar performance such as detection range, system resolution and accuracy, and potential ability of anti-interference. Based on the fuzzy function theory and MIMO formation parameters, this paper analyzes the time-frequency-space 3D fuzzy function of DDMA-LFM waveform set, and analyzes the relationship between the waveform set and array flow pattern and radar detection performance in range-velocity-azimuth three dimensions. Combing the transmitted waveform set and the radar array flow pattern to derive the three-dimensional time delay-Doppler frequency-space orientation MIMO ambiguity function. Through formula analysis and simulation results, it is concluded that the continuity of the waveform is ensured and the design process is simplified.. In the process of designing the orthogonal DDMA-LFM waveform parameters of the Multiple-Input Multiple-Output (MIMO) radar system, there is no need to optimize the coding. What’s more, the range (time delay) resolution performance and the spatial orientation can be Separated independently, and the Doppler frequency and the spatial orientation are mutually coupled but the radar resolution performance is not affected within the effective Doppler spectral width, meanwhile the interaction between the waveform set and the array can improve the angular resolution of certain orientations to a certain extent about 5°.

Abstract:For the sake of settle the disputes of algorithm poor optimization performance and slow training speed of gray wolf optimization algorithm, a new gray wolf optimization algorithm is proposed. In the initialization part of the algorithm, the reverse learning strategy is used to generate ordered gray wolf individuals, which validly ameliorates the convergence speed of the algorithm. The search ability of the algorithm is coordinated by designing a new nonlinear convergence factor and optimizing the individual location update strategy to reduce the probability of falling into local optimization. The elite selection retention strategy and tournament selection strategy are introduced to accelerate the population evolution and improve the convergence speed of the algorithm. The basic function test results and track planning simulation experiment verify that the new gray wolf optimization algorithm has strong astringency and high optimization accuracy, and the average track generation value spent by the algorithm is 19.9% less than that of gray wolf optimization algorithm.

Abstract:Convolutional operation only extracted local time-frequency information, and cannot effectively mine the relevant information between spectra. In order to solve this problem, a spectrum shift densenet was proposed. The module adopted structure of dense convolutional module, and the spectrum shift module was used to realize the information interaction between the spectra. It replaced the down-sampling operation between spectra and extracted the global feature from the spectrum. Meanwhile, it avoided the loss of information in the down-sampling process and further improved the quality of the spectrum feature maps. The proposed method was verified on two widely used dataset ESC10 and ESC50 respectively. The classification accuracy of ESC10 and ESC50 datasets is 96.00% and 88.75% respectively. Compared with the existing networks,the accuracy is improved by 2.1% and 2.25%. Comparde with convolutional neural networks based other methods, the proposed module can effectively mine more time-frequency information and has higher accuracy.

Abstract:Frequency estimation of sinusoidal signal is widely used in communication, radar and sonar, etc. Forasmuch as Accurate estimation of sinusoidal signal frequency, an frequency estimation method for sinusoidal signals based on quadrant discriminating and moving fitting is proposed. The period or phase step of the sampled signal is estimated preliminarily using the quadrant discriminating method. Then, the move sine fitting method and the estimated parameters are applied to processing of sequence data further. Finally, the signal period or frequency is calculated directly according to the fitted phase of corresponding sequence point. The proposed algorithm is compared with simulation accuracy of three existing interpolation algorithms based on discrete Fourier Transform (DFT), The simulation results show that when the design SNR is -5,0,5 dB, measuring frequency is about 11.432Hz with six discrete frequencies, sampling frequency is 2222.22Hz and data quantity is 5112, the mean root mean square error (RMSE) of frequency estimation of this method is respectively 2.048×10-3，1.290×10-3，0.870×10-3Hz. The mean of mean absolute error(MAE) is respectively 1.823×10-3，1.209×10-3，0.687×10-3Hz, which is lower than the errors of the other three algorithms. The proposed algorithm has higher accuracy and is closer to the Cramer-Rao lower bound (CRLB). The proposed algorithm is stable performance due to insensitive to the range of frequency.

Abstract:Aiming at the error problem in the high frequency band of the standard frequency when the traditional bilinear transformation method is used to design the frequency weighting filter in the sound level measurement, a frequency weighting design method based on the chaotic weighted particle swarm algorithm is adopted to realize the frequency weighting Optimize the design. First, analyze and study the error source of the frequency-weighted amplitude-frequency response, and then design the optimization object and fitness function of the frequency-weighting filter according to the analysis results, and construct a frequency-weighted digital filter optimization model based on chaotic weighted particle swarms , And seek the optimal solution. The simulation experiment shows that the optimization algorithm proposed in this paper has high accuracy and the error can be controlled within 0.005dB, which meets the design requirements of the first class sound level meter in the national standard, and It is suitable for frequency weighting in sound level measurement and filter design in other fields.

Abstract:Aiming at the problem of electromagnetic interference in the shielded cables of high-speed railway. On the basis of the analytical model of single-shielded transfer impedance. Considering the effect of the double- shielded on the DC impedance and small-hole inductance of the shielded layer, an analytical optimization model is proposed for the shielded cables of high-speed railway. Combined with the theory of transmission line and shielding efficiency, the quantitative relationship between shielding efficiency and transfer impedance is proposed, and the theoretical mathematical model of core wire subjected to conduction interference is given. Finally, the shielded cable of high-speed railway is taken as the research object for modeling and simulation, in order to analyze the disturbance situation of the cable. Comparing the simulation value and the calculated value, which verifies the correctness of the model. The error rate of the optimization model is less than 10%, which verifies the effectiveness of the model. The disturbance mechanism of shielded cable of high-speed railway is revealed.

Abstract:With the development of transportation and logistics, higher requirements are put forward for the time complexity and accuracy of path planning algorithm. Aiming at this problem, this paper proposes an improved intelligent path selection algorithm based on grid map which uses the direction information to optimize the algorithm. Based on the direction information of the beginning and end positions in the original map, the weight matrix and direction weight vector are defined as the basis for the selection of the next grid in the path, and reward and punishment measures are taken to improve the convergence speed of the algorithm. In addition, the algorithm also takes the optimization measures of redundant grid to improve the performance of the algorithm. The effectiveness of the algorithm is verified by the simulation of a typical storage AVG problem model. Simulation results show that the time required for the path length to converge to the optimal solution can be reduced by more than 10% compared with A * algorithm and traditional Dijkstra algorithm. The proposed intelligent path selection algorithm has lower time complexity and higher optimization efficiency as well.

Abstract:Although the traditional convolutional neural network has good application accuracy, its main defect is low efficiency. In order to solve this problem, the weak supervised learning algorithm is proposed. The existing weak supervised learning algorithm has less labeled training samples and ideal efficiency, but there is still a lack of high misclassification rate. In order to meet the requirements of high efficiency and high precision at the same time, this study combines weak supervision algorithm and convolutional neural network, a weak supervised convolutional neural network feature learning algorithm based on class space constraints is proposed. Firstly, the network model of the feature learning algorithm of weakly supervised convolutional neural network was established; secondly, by constraining the space, the labeled samples and unlabeled samples were connected to realize the feature space clustering; finally, the training sample data was used to realize the design of weak supervised convolutional neural network feature learning algorithm based on class space constraint. The results show that the proposed method has a misclassification rate of 5% and a classification time is no more than 0.4ms, which can better carry out feature learning.

Abstract:In order to guarantee the reconstruction visual quality while improving the reconstruction rate, this paper proposes a video super-resolution reconstruction algorithm (SPyGAN) based on spatial pyramid generative adversarial network, which uses a more lightweight spatial pyramid network structure SPyNet and a more efficient upsampling method based on TecoGAN, and can quickly reconstruct the high-frequency texture details of images. In this paper, we mainly improve the optical flow prediction network, image reconstruction module and loss function part of the generative adversarial network TecoGAN. Experimental results show that the algorithm has improved the mean values of PSNR and SSIM compared with TecoGAN, in addition to the reduction of the parameter amount to 53.86%, and the reconstruction rate is improved to 239%, which effectively improves the reconstruction rate of the model.

Abstract:The characters on the surface of IC chip mainly include the manufacturer's name and serial number. These characters are of great practical significance for chip manufacturing and application. For the detection of printed characters on the chip surface, a chip character recognition system is developed based on Halcon visual software development platform. Firstly, the gray value projection method is used to obtain the row and column coordinate segmentation points of the character region for character segmentation. Then, the shape matching technology is used to locate and correct the chip image to be detected, and the BP neural network classification algorithm is used to realize character recognition. Through the comparative experimental analysis of different algorithms, the experimental results show that the detection time of a single picture is 42ms, the segmentation accuracy of complete characters and defective characters is 100%, and the character recognition rate is 99.5%. The system can effectively, quickly and accurately recognize the characters on the surface of IC chip, and the detection accuracy meets the requirements.

Abstract:Automatic identification of chip identification codes is of great significance to avoid manual plug-in errors and improve sorting efficiency. In this paper, a chip surface identification system based on OCR technology is designed. The system consists of an industrial camera, a light source, a detection platform, a trigger device and a PC. The trigger device triggers the industrial camera to capture the chip picture. The OCR algorithm is used to automatically identify the chip surface identification code in the picture and extract the identification code for subsequent application. The application results show that the recognition time of a single image is about 300ms, and the recognition accuracy can reach more than 95%. Under the premise of ensuring certain accuracy, the work efficiency can be greatly improved, and it has good practical value.

Abstract:To solve the problem that the on-line detection of shape and position tolerance of parts in precision machining industry is not real-time and can not detect multiple parts at the same time, the image preprocessing process and measurement method of images collected by camera are improved by using machine vision technology, and an improved non-contact measurement algorithm based on CNN super-resolution reconstruction is proposed. Compared with other super-resolution reconstruction algorithms, the algorithm has the advantages of simple model, high precision and fast speed. It can take into account the measurement accuracy and efficiency under the condition of limited resources. In order to verify the reliability of the designed algorithm, a non-contact measurement system based on machine vision is designed. The experimental results show that the accuracy of the improved measurement method can be improved by at least 47.86% and 49.67% on average compared with the previous measurement method. The super-resolution algorithm on the basis of the resolution must, to the original acquisition of image super-resolution reconstruction after improve image resolution, measurement accuracy is increased by 60.38%, do not use the super-resolution reconstruction using the algorithm for multiple targets online synchronous measurement analysis, and the precision is not lower than with single parts under resolution precision.

Abstract:Affected by the poor working environment and other reasons, the fan blades often have defects such as cracks and pits. Aiming at the low accuracy of the current common target detection algorithms for the detection of small-size defects of the fan blades, a fan blade defect detection method based on the EfficientDet algorithm is proposed. . First collect image data and establish a wind turbine blade defect image data set in Pascal VOC format, and then improve the backbone feature extraction network in the EfficientDet algorithm to reduce the number of downsampling and adjust the effective feature layer to enhance the backbone feature extraction network for small-size defects Detection capability; At the same time, the multi-scale feature fusion capability of the fusion path enhancement algorithm is added to the feature fusion network. The algorithm uses FReLU as the activation function to achieve pixel-level spatial information modeling, and uses Mosaic data enhancement and Focal Loss loss function to increase small-size defect samples for Contribution of the detector. The test results on the established defect image data set of fan blades show that the improved algorithm model has an average category accuracy of 96.15%, which is 3.77% higher than the original EfficientDet, and the detection performance of small targets has been significantly improved.

Abstract:Aiming at the problem that the number of samples in few-shot learning is too small to represent the characteristics of data categories, a few-shot learning method combining prototype calibration data distribution is proposed. First, the embedded network is used to preprocess the image to obtain new class features, and the extracted new class features are processed by power transformation. Then, the prototype of the new sample is characterized by the weighting of the similarity of the base class, and the learned knowledge of the base class is fully utilized to reduce the deviation between the calculated prototype and the actual prototype. Finally, we sample from the uniform distribution between the features of the new class instance and its prototype representation. It generates a large amount of feature data to expand the support set of the new class. At the same time, we propose a method to change the boundary of the uniform distribution according to the number of samples. As a result, the samples are concentrated in areas with high confidence. The accuracy of 5-way 1-shot and 5-way 5-shot of our method are 68.94% and 84.75% on the miniImageNet dataset, respectively, and the accuracy on the CUB dataset are 81.75% and 91.88%, respectively, which are better than the best results of existing methods. The experimental results show that our method can effectively improve the model’s prediction accuracy in few-shot classification.

Abstract:Aiming at the mismatching problem of traditional semi-global matching (SGM) in high-resolution images with weak textures and disparity discontinuities, an SGM algorithm that takes into account the image segmentation information is proposed. In the cost calculation stage of this algorithm, the size of the matching window is first adaptively adjusted according to the image segmentation information, and the improved Census transform with different state information is used to calculate the initial cost, which solves the traditional algorithm's dependence on the center pixel of the Census transform window and reduces the matching time. In the cost aggregation stage, the image segmentation information is organically combined with the global energy function of the traditional SGM algorithm to improve the matching effect of the algorithm in weak texture and depth discontinuous regions. Finally, the optimized disparity map is obtained through left-right consistency detection and sub-pixel refinement. The proposed algorithm is verified by using the standard data of the middlebury platform. The experimental results show that the average mismatch rate is 4.54%. Compared with the traditional SGM algorithm and some improved algorithms, the proposed algorithm can obtain higher matching accuracy in the weak texture and discontinuous disparity areas of the image.

Abstract:Aiming at the low efficiency of traditional projectile image interpretation, an image automatic interpretation method based on improved edge detection has been proposed to obtain the coordinates of projectile feature points. Roberts template is used to calculate the image gradient at the sub-pixel level, and the maximum gradient in the neighborhood is found as the anchor point. The single pixel edge of the projectile is obtained by the Smart Routing method and the centroid coordinates of the projectile are obtained by moment theory. Using the resolution plate to test the improved edge detection method, and the projectile centroid in shadow image is obtained. Experimental results show that, compared with the traditional method, the improved algorithm has better effect on edge continuity and positioning accuracy, and the error of the projectile centroid’s abscissa is reduced from 6.7 % to 2.8%.

Abstract:The early warning of satellite is a necessary process and an important part of the early warning system, which is divided into double satellite detection and single satellite detection. The traditional single satellite detection method is based on the prior information of standard trajectory template, but the standard trajectory template information is difficult to obtain.In this paper, a missile trajectory estimation model by early warning satellite based on direction constraint is proposed aiming at the problem of trajectory estimation of ballistic missile by early warning satellite. The trajectory of ballistic missile is estimated and predicted by using the detection data of the missile boost phase of single satellite and double satellite detection respectively , and the quantitative calculation is carried out. During single satellite detection, the smaller the angle between the satellite position and the missile plane, the greater the error. When the angle reaches more than 5 °, the error is small, which can basically meet the position guidance requirements of long-range early warning radar. Double double star detection, The prediction error is further reduced and the accuracy is further increased during.The above simulation results show that the trajectory estimation accuracy of this method can provide enough target guidance accuracy for ground-based early warning radar.

Abstract:In order to ensure the normal operation of the high-voltage power supply system, the high-voltage power battery pack needs to meet insulation safety requirements during normal operation. An improved method for on-line detection of insulation resistance and fault point location is proposed to solve the problems of low accuracy and poor reliability of insulation resistance detection of power battery packs, and inability to locate the internal insulation fault location of the battery pack. This method combines balanced bridge and unbalanced bridge to measure, improves the calculation method of insulation resistance, can quickly and accurately realize the insulation resistance detection of power battery pack, and can estimate the location of insulation fault inside the battery pack. According to the improved measurement principle, the hardware and software of the detection system are designed using microcontrollers, optocoupler relays, differential analog-to-digital converters, digital signal isolation, isolated power supplies, etc., to achieve insulation resistance detection and fault location. Experimental test show that the improved detection method and detection system can accurately detect the insulation resistance value, the detection error is within 5%, and the insulation fault point is located accurately, which provides a practical method for the insulation resistance detection and fault location of the power battery pack.

Abstract:Shanghai soft X-ray free electron laser facility (SXFEL) is the first fourth generation light source in China, and its shortest wavelength can reach 2 nm. The facility consists of an injector including an S-band photocathode Radio Frequency (RF) gun, two S-band accelerating sections and one main accelerating section including six C-band accelerating sections. In the low level radio frequency (LLRF) system, C-band microwave system operates at 5712 MHz. C-band microwave signals are mixed with the 5686.556 MHZ local oscillator (LO) signal on the RF front-end board and down converted to 26.444 MHz intermediate frequency (IF) signal, and then sampled by 105.778 MHz clock signal. According to the physical design requirements of SXFEL, DWC8VM1 board based on MTCA platform is selected as the RF front-end board of LLRF system. In this paper, the parameters of the RF front-end such as isolation between RF input ports, input and output linearity, amplitude and phase stability and Additional phase noise of up conversion channel were experimentally tested. The experimental results show that input return loss is more than 8.63 dB, channel-to-channel crosstalk is more 63 dB and the additional phase jitter integral of up conversion channel is 2 fs. The performance of this RF front-end (DWC8VM1) can meet the theoretical requirement of the SXFEL.

Abstract:Aiming at the problem that the traditional convolutional neural network has low accuracy and poor performance in feature extraction in gearbox fault diagnosis, a SincNet combined with attention mechanism method for gearbox fault diagnosis was proposed. First, use the parameterized Sinc function to design the filter and obtain the Sinc convolutional layer, Sinc convolutional layer replace the first convolutional layer of traditional CNN to construct the SincNet network structure, Extract the characteristic information of the input data. Then, combined attention Mechanism with Softmax enhances characteristic information. Finally, the gearbox fault data set was used to verify the proposed method. The results show that the average diagnostic accuracy of the proposed method is 99.68%, which is higher than that of the comparison method. In addition, the method can accurately locate the recognition information in the input data and better understand the feature extraction process of neural network through the visual analysis of the feature map, which provides a reference for the feature extraction process of mechanical vibration signals.

Abstract:With the continuous renewal of light source products, people are paying more and more attention to healthy lighting. The detection of lighting quality is an important guarantee of healthy lighting, especially the detection of lighting quality parameters such as intensity, flicker frequency and flicker proportion. Aiming at health lighting, design a kind of lighting lamps and lanterns lighting quality detector based on STM32, silicon photocell chosen as the photoelectric sensor, through the current voltage conversion and low pass filtering to extract the signal intensity of the light, through band-pass filter to extract the stroboscopic signal, using the stroboscopic signal envelope detection circuit as a picture, using the stroboscopic signal frequency discriminator circuit. STM32F103RCT6 microcontroller is used as the core processor to analyze and process the information, and obtain the quality parameters such as light intensity, flicker frequency and flicker proportion. The main advantages of the system are that it can detect the multi-parameter quality of lighting lamps online, with simple structure, small size, low cost and good stability. It can be used to detect the lighting quality of household lighting lamps, and help people screen out better lamps and achieve healthy lighting. After the experiment, the system can accurately detect whether the household lighting on the market is qualified, the detection accuracy can reach more than 99%. It can detect whether the light source is healthy and provide a reference method for detecting whether the light source is safe.

Abstract:Atomic force microscopy (AFM) is used to characterize the properties of material in microscopic scale, which has a board application in scientific research. Most of the commercial atomic force microscopes work in amplitude modulation mode (AM-AFM), and the resolution of imaging is limited to quality factor Q. In this paper, we developed a frequency modulated atomic force microscope (FM-AFM) to improve resolution, and analyzed the important factors affecting the system resolution. In addition, we designed a vacuum system to improve the quality factor Q of the probe cantilever. For comparing the superiority in sensitivity and measurement effect of FM-AFM in vacuum environment with in atmospheric environment, the system performance of FM-AFM in vacuum environment were studied. The results show that the power spectral density of the system is 10mHz/√Hz, the system sensitivity is less than 0.02Hz, and the quality factor Q of the probe cantilever is 14212, which breaks the limit of Q in atomic force microscope.