As a challenging visual task，single target tracking plays an important role in the industrial and military fields. From the traditional method with image features as the core to the deep learning method centered on network architecture design，it has shown a lot of research value. The development process of single target tracking is summarized in this paper. Some popular data sets are introduced. The correlation filter methods based on speed and the deep learning methods based on high accuracy are taken as the main principles. The design ideas of some baseline methods and high performance methods are researched. Finally，the methods of various architectures are summarized，and the subsequent research trends are forecasted.

As a high-precision surface shape measurement method，the measurement accuracy of deflection technique not only depends on the calibration accuracy of system parameters，but also is affected by the display surface shape. Especially for the commonly used deflection technique measurement system based on flat mirror reflection model to achieve system calibration，the display surface shape will directly affect the calibration accuracy of system parameters. In order to study the influence of the display surface shape on the measurement accuracy of s stitching deflectometry，this paper firstly presets the display surface shape of different shape variables with the help of the proposed algorithm，and then analyzes its influence on the system calibration accuracy and measurement accuracy. The results show that the display surface shape will reduce the calibration accuracy of system parameters，in addition，an amount of low and middle order errors is introduced into the surface shape measurement results of the components under test. Finally，by comparing with the experimental results，the conclusion obtained by the algorithm is further verified. This study provides a quantitative calculation and analysis method for the measurement error analysis of stitching deflectometry system.

In order to solve the problem that the edge detection cannot be realized a prior and the edge extraction efficiency is too low，a more efficient subpixel level edge detection method based on Fourier single pixel imaging is proposed. Combined with fast Fourier single pixel imaging，this method reduces the number of phase shift steps of the image algorithm，and realizes the three-step phase shift and two-step phase shift edge detection respectively on the basis of the original four-step phase shift. The improvement of the algorithm can expand the spectrum width of edge extraction under the same number of samples，so as to improve the efficiency of edge extraction. The numerical simulation results show that， compared with the four-step phase shift subpixel level edge detection，the peak signal-to-noise ratio（PSNR）increases 2.27 dB in the sampling range of 655~13 100 times under the condition of no noise，and the two-step phase shift method can obtain a higher edge extraction quality than the four-step phase shift when the noise floating ratio is lower than 0.054. The method proposed in this paper can improve the efficiency of edge extraction to a certain extent，and promote the technology crossover and application development of the single pixel imaging field and image processing direction.

Transient fracture is the main performance index of precision conductive slip ring. It has the characteristics of low occurrenceprobability，short timeanddifficult todetect. However，it iseasy tocauseuncertain impact on the system， which directly related to the reliability of system signal transmission. In this paper，the cause of attitude jump during the modulation of inertial navigation equipment turntable is analyzed. The instantaneous breaking phenomenon of the conductiveslipringisrepeatedbythedetectionsystem，anditpoints outthatsmallforeign matterenteringtheringgroove willinevitablylead tothetransientbreakingphenomenonoftheconductiveslipringwiththecharacteristicsofsymmetrical doublecontactofbrushwireandirreversibledamageoftheconductiveslipring.

In order to solve the problem that the spatial transient magnetic field of the secondary equipment of the UHVDC converter station is not clear after being interfered by radio，the research on the monitoring technology of the transient magnetic field of the secondary equipment of the UHVDC converter station is proposed. Based on the data acquisition program of the space transient magnetic field of the secondary equipment of the converter station，according to the schematic diagram of electromagnetic radiation，the electric field intensity and electromagnetic radiation intensity of the secondary equipment space of the converter station are calculated. The switching arc is simplified into a straight wire，and the space transient of the secondary equipment is calculated. Combined with the space transient magnetic field monitoring algorithm of the secondary equipment of the converter station，the space transient magnetic field monitoring of the secondary equipment of the converter station is realized. The test results show that the method in this paper can effectively monitor the space transient magnetic field strength of the secondary equipment，and the radio interference level of monitoring the transient magnetic field is between 10~25 dB· μV/m，which meets the requirements of the standard interference level.化等。

One of the key technologies in automatic digital microscopy is autofocus. In order to improve the speed of focusing，more and more deep learning methods are being introduced for focus prediction of single-frame images. However，almost all networks believe that their output is necessarily correct，even in the face of unknown samples when the output error results will not include any warning. In this paper，a Bayesian convolutional neural network is proposed to predict the defocus distance from a single image and obtain the uncertainty estimation of the focus prediction results. In addition，uncertainty is proposed to measure the validity of the results，and the focus prediction results are filtered by setting the uncertainty threshold. The proposed method is tested on a large open-source dataset. Experimental results show that the network model proposed in this paper can output higher uncertainty for unknown samples，and the established screening mechanism can effectively reduce the prediction error of the model for unknown samples by eliminating some error results. The model achieved a final error range of 0.37 ± 0.46 μm and 0.83 ± 1.17 μm on two samples on the public data set，which is better than 0.40 ± 0.66 μm and 1.08 ± 1.78 μm before screening.

Aiming at the low detection rate of black polar ship targets（BPST）in visible remote sensing images（VRSI） taken by remote sensing satellites and high-altitude UAV at sea，a ship target detection method based on dark primary color prior and constant false-alarm rate principle is proposed in this paper. In this method，the target image is preprocessed to obtain its dark channel image（DCI）. The DCI is used as the input of 2-D-CFAR detection method，and the local statistical gray mean and variance are used as target detection characteristics. The suspected targets’locations are completed by selecting appropriate target determination threshold and target spatial distribution information. Finally，the suspected targets’ sets are verified by using the established ship target feature vector to remove false alarms and output the final target detection results. Experimental results show that this method has a more accurate detection effect on ship targets in VRSI，with a total detection rate of 96.07%. It also acts as a“magnifying glass”for BPST which are difficult to be distinguished by human visual，with a detection rate of 86.71%. It greatly improves the detection rate of BPST in VRSI，which has certain guiding significance to optimize the innovation of black polar target detection method，and also provides a new idea for the application of dark channel prior principle.

Aiming at the actual image brightness/contrast in fog environment，an overall gray level tensile and local contrast enhancement algorithm to improve the image brightness and contrast is proposed in this paper. The image removal algorithm based on dark primary color prior is used to remove the haze effect often encountered in video surveillance. In order to eliminate the blocking artifact，the image is divided into the smallest blocks，that is，the dark primary color is extracted for each pixel，and the adjacent similarity principle is used to correct the dark primary color. MATLAB simulation shows that the improved algorithm can remove haze from the image. Finally，the software development of DaVinci DM6467 based image enhancement algorithm is completed，and the output，switching and image enhancement of 4-channel video are realized. Enhanced image，its SSIM indicator can be increased by more than 50%. The system can effectively remove the effects of haze on the image to meet the needs of image dehaze enhancement.

In order to achieve the big image plane，small throw ratio，high definition picture requirements，a catadioptric ultra-short focal projection lens is designed. According to the requirements of performance indicators，the inverse telescopic system is selected，and the initial structure is obtained by the scaling method. The mirror coordinate data is obtained by calculation from the relationship between the light rays of different fields of view and the height of the image plane at the exit pupil position，and the mirror surface type is obtained by fitting. The image quality is evaluated by using point plot， field curve/distortion and modulation transfer function（MTF）curve. Finally，an object-space telecentric system with a focal length of 5.45 mm and a throw ratio of 0.4 is obtained. When the projection distance is 720 mm，it projects an 80 inch screen with a resolution of 1 080 p，a field of view of 136°，and a magnification of about 125 times. The RMS radius of the column diagram is less than 750 μm，the distortion is less than 0.2%，the relative illuminance is above 96%，and the MTF curve amplitude is greater than 0.3 at 0.54 lp/mm. Tolerance analysis shows that the system design is reasonable and the imaging quality is good. Compared with other catadioptric ultrashort focal projection lenses，the system can ensure the imaging quality without using aspherical surface in the refraction part，which effectively reduces the difficulty of processing and assembly.

In order to simulate the real shooting scene and reduce the cost of military shooting training，a laser shooting optical system based on erbium glass laser is designed. The system uses two-stage laser beam expansion to compress the divergence angle of the output beam of the erbium glass laser twice and ensure the output of the beam with the minimum divergence angle. It improves the human eye safety during laser shooting.The divergence angle of the output beam of the erbium glass laser is 15mrad and the diameter of the output spot is ?0.3 mm，the laser beam expansion rate is 300 times， and the beam divergence angle of the system is 0.035 mrad. 1km away from the optical system outlet，the laser spot diameter is about ?120 mm，which can meet the requirements of kilometer shooting training and break through the distance limit of the conventional laser shooting system. The total length of the designed optical system is 364 mm，the structure is simple，it is convenient for engineering，and the application prospect is broad.

In order to improve the low power efficiency of DCO-OFDM system and reduce the high complexity caused by high-order signal modulation，subcarrier index-power modulated with superposition multiplexing（SIPM-SPM）is applied to asymmetrically clipped optical orthogonal frequency division multiplexing（ACO-OFDM）system. The proposed signal transmission technique adopts SPM to add two complex data encoded by 8PSK and QPSK for the high power subcarriers， which further increases the transmission capacity and avoids the power loss caused by DC bias in ACO-OFDM system. Simulation results that in ACO-OFDM system，compared with SIPM（8-PSK/QPSK），the signal transmission capacity of SIPM-SPM is increased by 28.6%. Compared with SIPM（32-PSK/QPSK）with the same signal transmission capacity， when the bit error rate is 1.0 × 10-3，the signal-to-noise ratio gain is > 2.8 dB，the dispersion tolerance is increased by > 560 ps/nm，and the optical launch power dynamic range is improved by > 6.7 dB. It is shown that the combination of SIPMSPM modulation and ACO-OFDM system improves the noise resistance performance，high dispersion tolerance and reduces the demand for optical launch power.

Infrared temperature measurement technology is a common equipment fault detection method in power industry， which has important application in GIS equipment fault diagnosis. To improve the accuracy of temperature measurement，an infrared temperature measurement model is established and an SF6 transmittance correction method is proposed. First，an infrared temperature measurement model is established based on the thermal radiation theory，and its simplified model under different conditions is presented. Then，considering that the SF6 transmittance will influence the model output，and the correction method of SF6 transmittance under different conditions is presented to improve the temperature measurement accuracy. Finally，some experiments are carried out to verify the effectiveness of the established infrared temperature measurement model and the proposed SF6 transmittance correction method. The experimental results show that，compared with the method without SF6 transmittance correction，the temperature measurement accuracy is improved after using the proposed method，and the error is reduced by up to 66.7%. The proposed method lays the foundation for power grid fault detection and monitoring，and has broad application prospects in power grid fault detection and security operations.

Aiming at the problems of temperature change and tower tilt in OPGW transmission line monitoring，a wavelength division multiplexing multi-parameter monitoring array based on fiber Bragg grating is designed. The mathematical relationship between grating strain and temperature and tilt is deduced from the theoretical model，and then the tower model is imported into CMOSOL Multiphysics for qualitative analysis. Finally，the temperature and tilt test are carried out. The test results show that the tower temperature monitoring resolution based on fiber grating can reach 0.1 ℃. The angle correlation coefficient K value of the inclination sensor is 0.0501 nm/° ，which proves that the proposed monitoring array has a good monitoring application prospect in the power industry.

Laser speckle can seriously affect the quality of images，and can reduce the image contrast and resolution. To suppress laser speckle，based on the visual transient characteristics of the human eye and the principle of diffuse scatter speckle suppression，a small-sized laser speckle suppression device is constructed by making use of circulating colloidal solution. The device has a small volume and can effectively suppress speckle，achieving inhibition effects where speckle contrast is 5% lower than the speckle perception limit of the human eye and transmittance of light is higher than 65%. At the same time，this device has obvious advantages in reducing noise and price compared to vibration diffusers，and has certain practical promotion value.

The mechanical energy loss of the hemispherical resonator includes intrinsic loss of the material，thermoelastic damping loss，air damping loss，supporting loss and surface loss，which determines the Q factor of the hemispherical resonator. Through analyzing the mathematical model of the energy loss mechanism，the dominant errors affecting the hemispherical resonator are material properties，surface roughness，form and position accuracy. To solve the problem of processing precision and engineering applications，a processing technology to achieve the high precision hemispherical resonator is proposed. After optimizing of process parameters of precision grinding，precision polishing and chemical treatment，the surface roundness error of the hemispherical resonator is less than 0.25 μm，the roughness is less than 15 nm and the quality factor is greater than 5.2×107. The research in this paper may be have important guiding significance for high precision hemispherical resonator and hemispherical resonator gyroscope engineering applications.

It is important for the strap down inertial navigation system to keep the zero-bias constant during the long working time of the RLG. The main factors affecting zero bias of the RLG are introdured in this paper. The mechanism of the zero-bias drift during the long working time of the RLG is analyzed. Experimental results show that when the laser gyro works continuously for one month，the zero-bias drift is about 0.0053°/h，and the positive counterclockwise beam loss difference changes by about 8%. The effect of loss difference on zero-bias of laser gyro is long and slow. In order to reduce the zero drift of laser gyro in long-term operation，the long-term stability of the loss difference of forward and retrograde waves should be improved.

According to the precise positioning requirement of INS/BDS integrated navigation system in short-term occlusion scenes of high dynamic satellite signal，the BDS/INS tightly coupled filter method based on sequential algorithm is studied in this paper. The BDS/INS tightly integrated filter equation is introduced first. Besides，the method based on sequential algorithm is derived in detail，and the corresponding theoretical analysis is carried out. In the end，the filtering performance of proposed method is simulated and analyzed by signal simulator. The results of corresponding simulation indicate that the algorithm proposed can reduce the computational complexity of the system effectively without losing the filtering precision. Moreover，the applicability of the INS/BDS integrated navigation system is enhanced in short-term occlusion scenes of high dynamic satellite signal.short-term occlusion

Gravity/inertial matching navigation is an integrated navigation method which uses the earth gravity field information to correct the inertial position matching. In matching navigation，the spatial resolution error is caused by the spatial resolution difference between gravity measurement information and gravity database information. In this paper，a spatial resolution synchronization technique is proposed to reduce the spatial unsynchronization error in matching navigation. The spatial resolution synchronization technique calculates the spatial resolution of the gravity measurement information through the carrier motion information and the parameters of the gravity measurement algorithm filtering model，and reduces the resolution of the gravity database through the two-dimensional filtering. The information of gravity measurement is synchronized with the information of gravity database. After the spatial resolution of the database is modified，the matching navigation algorithm is used for matching location. The location accuracy of the improved matching navigation algorithm is 1.1 nmile higher than that of the traditional matching algorithm. The spatial resolution synchronization technology can improve the positioning accuracy of gravity-matched navigation，which will be helpful for the development of gravity-matched navigation equipment in the future.

Hemispherical resonator gyro has the advantages of simple structure，radiation and shock resistance，long life and high reliability，among which signal drive and readout circuit is the key to obtain high precision gyro signal. The unit circuits that affect the performance index of the gyroscope，such as signal readout circuit，excitation circuit and power supply circuit，are studied and optimized. Based on weak signal processing and filtering technology，the method to improve the detection voltage stability is studied. Tina-TI software is used to simulate the performance of the designed circuit，and the overall circuit design is realized with a gain of 40 dB and noise of 4.81 μV/ Hz. A gyroscope prototype is built to test the readout circuit. According to the experimental results，the evolution trend of the output signal under the action of different superimposed excitation signals of HRG is given. The results show that the amplitude of the output signal increases with the increase of DC and AC voltage. When the voltage of excitation signal is constant，the frequency change of AC signal will also affect the amplitude of the output signal of hemispherical resonant gyro. Therefore，the study of the influence of the excitation signal of the hemispherical resonator gyro on the output signal provides the direction for the design of the driving circuit of the hemispherical resonator gyro.