For the detection of surface cracks and hidden cracks in ferrite inductorthe traditional machine vision method has some problemssuch as fuzzy image and low recognition rate. an inductor crack detection method based on laser scanning thermal imaging technology is proposed. The thermal imager is used to capture the temperature change of the sample surfaceand the image is synthesized based on the second-order differential algorithm.The contour extraction algorithm is used to remove the interference of the peripheral contour and filter the uniform image with the second largest valueand the edge detection is used to extract the crack. Finallyusing the crack area and aspect ratio and hog features as input featuressupport vector machine is used for image recognition. The results show that all the cracks can be imaged correctlyand the correct rate of crack recognition is 98.5%which is better than other detection algorithms.

Auto registration of optical and Synthetic Aperture Radar (SAR) images has always been a difficult problem because of the significant geometric and radiation differences between the images. Firstlythe phase congruency model with radiation invariance is introducedand the direction and scale information of phase consistency are calculated by using indexed map. The directional index map is formed by the direction feature of phase congruencyand the scale index map is formed by the scale feature of phase congruency. Local Phase Congruency Statistics (LPCS)a local descriptor for the imagesis established. FinallyLPCS is used to obtain the points of the same name in optical and SAR imagesand affine transformation model is used to achieve registration. Experimental results show that the proposed method is highly adaptive to the radiation differences between optical and SAR imagesand the registration accuracy is high.

Remote sensing image registration is a research of significance in the field of image processing. Large rotation difference between multimodal images seriously affects the improvement of the final quality of image matching. To solve this problema rotation difference correction method based on double branch neural network is proposed. Firstlywith the calculation of eigenvalue and orientation of phase congruency informationthe rotation feature vector of the image is extractednamed RVPC. After thata double branch neural network R-FCN is constructed and fully trained to predict the rotation difference angle between images. Furthermoretwo RVPC vectors are input into the network and we get an output prediction vector and then calculate the prediction angle. Finallythe image is affine corrected based on the angle. On public data set SEN1-2the training accuracy of network R-FCN reaches 98.17%.

The inverse problem of fluorescence molecular tomography is seriously ill-posed and very sensitive to noise. In order to improve the accuracy of targeted fluorophorea reconstruction method is proposed combining a low rank matrix completion algorithm with manifold regularization to suppress noise. More priori information is utilized to improve the reconstruction results of inverse problem. In the proposed methodboth the low-rank attribute of the observation matrix and the target aggregation of the reconstructed light source were used to improve the accuracy of the inverse problem reconstruction. The low rank matrix completion algorithm was employed to suppress the background fluorescence; the manifold regular model was employed to get a more aggregate and accurate solution. Finallyaccuracy of the reconstruction results were improved by this mean. To validate the proposed methodnumerical experiments were conducted. The experiment results demonstrated that the inverse problem reconstruction results were improved greatly in aggregation and accuracy.

Fluorescence lidar technology is one of the main methods for long-distance detection of organic aerosols. It is widely used in the detectionearly warningidentification and tracking of organic aerosols. In order to detect organic aerosols and prevent hazards as soon as possibleit is of great significance to improve the detection time resolution and early warning speed of fluorescent lidar system. According to the requirements of high time resolution and fast response of the fluorescent lidara design scheme of data acquisition system is proposed. The system divides the lidar signal detection process into data production and data consumption based on the producer-consumer model. Combined with the software multithreading technologythe coupling degree of the two modules is reduced and the parallelism degree of the two modules is increased. Thereforethe detection cycle of the fluorescent lidar is shortened effectively. The detection results show that when the multi-pulse accumulation acquisition period is 100ms and the number of single trigger sampling points is 20000the improvement rate of the lidar detection time resolution reaches 40.51%and it has good consistency with the theoretical formula. The system can be applied to the high time resolution data acquisition of fluorescence lidar.

Ultraviolet microscopic measuring system has been widely used in bioscience and semiconductor industrythe tube lens is an important part of the infinite conjugate microscopic system. For the contradiction between the requirement of zoom function and the accuracy of focal lengththe design of a near ultraviolet-visible triple focal length tube lens system with high resolutionwide field of view (FOV) and low distortion was introducedwhich has working wavelength range of 355~500nm. The structure is special in that it is composed of common-path formal component and parfocal rear componentthe focal length can be changed accurately by replacing the rear component. The tube lens system was designed and optimized by ZEMAX and their aberration and MTF were analyzed. The design result of this system shows that the MTF of each FOV all approaches the diffraction limit at 70lp/mm in three focal length modes. The distortion is less than 0.5% with high relative illuminance of image plane. The system can satisfy the design index well under reasonable tolerance parameters and can be adapted to varieties of ultraviolet microscopic objective lenses for accurate microscopic measuring imaging.

Dynamic display technique based on microlens array and micropattern array can realize three-dimensional image with stereo effect. A design and fabrication method for microlens array and micropattern array integrated on polymer films was introduced. Design formulas for the periodorientation and displacement of dynamic pattern were derived by the proposed moire geometric method. Stereo effects of floatingsinking and relative shifting can be achieved by the proposed method. Aperiodic imaging mode for the dynamic pattern was studied. The aperiodic image that can appear and disappear with view angle varying is obtained. Dynamic displaying films were fabricated using lithography and embossing processwhose periodorientation and displacement were measuredshowing their maximum deviations from the theoretical values were 11%16% and 14%respectively. The proposed method could be applied for estimating the performance and fabrication tolerance of dynamic displaying films effectively.

By using the thermal solution methodthe surface of the inverted chip was coated with a remote layer and a fluorescent adhesive film layer doped with nanoparticles of a certain concentrationand the gradient double-layer structure white light LED module samples were prepared.The photochromic properties of the samples were tested and the mechanism was analyzed. The results showed that adding remote layer between the chip and fluorescent adhesive film increased the light quantum number at the center normal line by 32.35% compared with the uncoated remote layerthe slope of the fitting line tends to be horizontaland the uniformity of photon spatial distribution is improved obviously. With the increase of SiO2 doping concentrationthe color temperature distribution curve gradually becomes flat. The average color temperature decreases from 6199.66k when the doping concentration is 0% to 5103.24ka decrease of 21.48%. The measured value of light flux first increased and then decreased with the increase of SiO2 doping concentration. When the doping concentration was 0.6%the maximum value reached 181mlmwhich increased by 9.70% compared with the undoped SiO2 particles. The efficiency of photon extraction is improved by means of gradient decreasing refractive index between the interfaces of double-layer white light LED moduleand the consistency of space color is improved obviouslyproviding a reference for improving the quality of white light LED module.

In order to develop an anti-infrared and 1064nm laser filter wit efficient electromagnetic shielding functionbased on the induction theory of metal film and the interference principle of multilayerthe influence of thickness error of Ag film on the spectral properties of filter is discussed. The dielectric films were deposited by ion beam assisted depositionand the metal Ag film was deposited by ion beam sputtering. The thickness of Ag film can be precisely monitored by means of enlarging the film thickness control errors. And the oxidation of Ag film can be avoided. The transmittance of the filter at 1064nm is over 88%the reflectivity of mid-long wave infrared band is over 90%and the shielding effectiveness of the filter at 18~36GHz is over 23dB. The filter has good shielding function of mid-long wave infrared and electromagnetic.

Based on the relationship between the refractive index and the pressure of gasthe dynamic pressure of gas can be measured non-contact by laser interferometer. At the same timethe influence of temperature on the measurement of gas dynamic pressure by laser interferometry was studied. The virial expansion of gas equation of state was carried out from the angle of quantum mechanicsand the model of gas pressure and refractive index was established. Based on the latest correction of Edlen equationthe influence of temperature on the measurement of gas dynamic pressure by laser interferometry is explored. The results show that when the pressure is constant in the low pressure rangethe temperature change in the range of -20~80℃ is inversely proportional to the refractive index of the gasand the change of the refractive index is about 10-6/℃. The temperature change of 1℃ is equivalent to the pressure 311.47Pa. The temperature change has a great influence on the measurement of gas at low pressure. It should be ensured that the temperature control is better than ±0.05℃so as to meet the requirements of dynamic pressure measurement by laser interferometry.

Based on the continuous observation data of lidarthe aerosol backscattering coefficient (355/532/1064 nm)extinction coefficient (532/607 nm)depolarization ratio (532p/532s)lidar ratio (532nm) and backscatter-related A°ngstrm exponent (355/532nm and 532/1064nm) are retieved. Thenthe aerosol optical properties of three different pollution episodes (air pollution/polluted dust/pure dust) in Beijing in October 2019 are analyzed. Results show that the air pollution aerosol depolarization ratio at 532nm (A°ngstrm exponent between 355 and 532nm) is 0.10±0.02(1.2±0.19)and the lidar ratio at 532nm (43±7sr) is lower than that of typical urban pollution aerosolwhich is probably affected by hygroscopic growth of water-soluble aerosolssuch as nitrate or the formation of secondary organic aerosols. In the polluted dust episodethe depolarization ratio (A°ngstrm exponent) is 0.19±0.03 (1.0±0.35)and the lidar ratio is 51±7sr. In the pure dust episodecompared with the formerthe depolarization ratio (0.25±0.03) is largerthe A°ngstrm exponent (0.11±0.44) is smallerand the lidar ratio is 40±4sr.

In order to measure the velocity field of high-heat and high-speed jet flow quickly and accuratelya measurement method of velocity field for combustion exhaust plume based on particle tracing is proposed. By actively adding metal particles with gray body radiation characteristics to the exhaust plumehigh-speed imaging is carried out by adopting appropriate observation parametersand the maximum cross-correlation degree of particle images is evaluated by combining gray distribution cross-correlationso as to obtain the velocity of flow field. The coaxial jet flame system is applied to the experimental research of the velocity measurement. The results show that the relative deviation between the measured value and the one-dimensional theoretical value at flow velocity of 90~140m·s-1 is less than 5%and the relative deviation of flow velocity measured at particle blowing speed of 0.02~0.1m·s-1 is less than 0.35%. It is verified that the method is suitable for similar application scenarios of velocity field measurement for exhaust plume in aero-engine test.

A general detection device was designed based on the traditional scanning pentaprism and image processing methodwhich can quickly and accurately detect the parallelism of the light from collimator with aperture less than 1 meter. The influence of pentaprism attitude on device’s angle measurement accuracy was analyzed in detailand the beam parallelism of 1-meter level collimator with 30-meter focal length in the laboratorythen the focal plane according was adjusted to the measurement results. The final collimator defocus is controlled within 4 mmand the angle measurement accuracy of the device can reach within 1 ".

The traditional pixel-level edge detection algorithm Canny algorithm is improved by replacing the Gaussian filter with a guided filterwhich better preserves the sharpness of edges on the basis of noise reduction and smoothingwhile designing the algorithm to adaptively obtain high and low thresholds suitable for images. An improved subpixel edge detection method based on the inverse tangent function model is proposedwhich calculates the subpixel edge position by the theory of equal area based on the centrosymmetric property of the inverse tangent functionwhich greatly reduces the computation volume and computation time compared with the traditional method. By fitting a circle with 3 times least squares to the top picture of the springthe information of the inner diameter and outer diameter dimensions of the spring is quickly and simultaneously obtained. The experimental results show that the average computation time of the proposed method is reduced to less than 1.7% while ensuring the detection accuracy compared with the traditional inverse tangent sub-pixel fitting method.

In order to improve the detection ability of infrared dim small target in complex environmentby introducing the human visual attention mechanisma weak small target detection algorithm based on difference histogram of oriented gradients coupling saliency mapping is designed. Considering the intensity difference between the real object and its neighborhoodthe DFT was used to calculate the amplitude and phase spectrum of the infrared image for calculating the spectrum residual between themand the saliency mapping was outputed by combinning with Gaussian filtering method to effectively highlight the salient region. The difference between the gradient amplitude and the gradient direction of the target and the background area was analyzedand the difference histogram of oriented gradients of the infrared image was calculated to fully suppress the background clutter and noise. And the fusion feature mapping corresponding to the image was formed by combining saliency mapping and multi-scale difference histogram. Finallythe adaptive threshold segmentation method was introduced to accurately locate the real target from the fusion feature map. Multiple sets of test data show thatcompared with the existing infrared target detection technologythe proposed algorithm can better locate small and weak targetsand have an ideal ROC curve.

Infrared enhancement is one of the effective methods to improve the quality of infrared imaging and highlight target information. The visual saliency map is calculated in a local window to help achieve the enhancement of the infrared image. Firstlysaliency is measured using center-surrounding pixel gray weight distance method within a local image region. Secondlysaliency map is obtained to simulate human visual system to assign weights to different pixel areas. Finallyinfrared image enhancement is achieved based on saliency map extraction. Several sets of infrared images are selected for experimentsand both subjective and objective evaluation methods are adopted to evaluate multiple enhancement methods. Compared with other methodsthe experimental results show that the results of the proposed method has better visual effectscan highlight the details especially the target informationand can effectively achieve contrast enhancement between the target and the background.

In the situation that a vehicle is driving along the lane and there is a camera above the lane. The relationship between the vertical coordinates of the reference point on car body on the image plane of the camera and the driving distance of the vehicle are used to obtain the driving distance of the vehicle indirectlyso as to identify the vehicle speed. Under the condition of general lane cameras’ parameters (focal lengthangleheight)through theoretical analysis and experimental verificationit is concluded that the vertical coordinate of the reference point on car body on the image plane of the camera has an approximately linear relationship with the reciprocal of vehicle driving distance. By selecting a reference point at the front or rear of the car body and calibrating the lane camera on site a linear fitting formula of the vertical coordinate of the reference point on car body on the image plane and the reciprocal of the vehicle driving distance can be obtainedand then the vehicle speed can be measured. This method with easily operation and good measurement results can be used in the judicial authentication of vehicle speed for traffic accident.

Aiming at the problem that the traditional tracking algorithm is greatly affected by the confusion between background clutter and different target typesa multi-target tracking method using n-type sequential Monte Carlo PHD (SMC-PHD) filter is proposed. Firstlybased on the stochastic finite set theorya new extended probability hypothesis density (PHD) filter modelnamely n-type PhD filteris proposedand the updated PHD is derived from the probability generating function. Thenthe sequential Monte Carlo (SMC) method is used to generate random samples. Combined with N-type PHD filterN-type SMC-PHD filtering algorithm is proposed to reduce the confusion detection of different target types. Finallythe measurement driven SMC-PHD filter is used to distinguish the measured attributes by the gating methodso as to suppress the influence of clutter on the existing target tracking. Based on MATLAB simulation platformthe proposed algorithm is tested in MOT17 and VS-PETS' 2003 soccer video data sets. The results show that the average optimal sub mode allocation (OSPA) distance of the proposed method is significantly reducedand it has high reliability in the case of high clutter and more target types.

In order to improve the tracking accuracy of infrared small target under the interference of background clutter and imaging noisean infrared small target tracking method based on structural information modeling and discriminant sparsity is proposed. Firstlythe small target signal is sparse decomposed in the generalized Gaussian target super complete dictionary to extract the spatial structure information of the small target from the infrared image which is corrupted by noise and clutter. Thenthe particle filter algorithm is improvedand the transfer constrained particle filter tracking algorithm is designed to improve the sampling probability of particles. Finallyin the framework of transfer constrained particle filterthe sparse coefficients of candidate targets are calculated based on discriminant sparse representation and L1 norm minimization framework to achieve small target tracking. Experimental results based on various infrared sequences show that the proposed method can track small target stably under the interference of clutter and noiseand its center erroroverlap rate and average video playback frame rate are 3 pixel0.7 and 0.632fpsrespectivelywhich are better than other comparison methods and have strong robustness.

Aiming at the problem of low detection accuracy of small target in infrared target detectionan improved Average Absolute Gray Difference (AAGD) algorithm is proposed to detect small infrared targets. Firstlyin view of the shortcomings of aagd algorithmon the basis of itthe gray and saliency features are fused and used in the kernel correlation filter to solve the problem of simple features and less information of infrared target. Thenan adaptive double sliding window is proposedwhich adjusts the shape of the aggregation window and the weight of pixels according to different regions to realize the matching of maneuvering targets near the background of high-intensity structure and improve the accuracy of small target detection. FinallyExperimental demonstration of the proposed method based on the MATLAB simulation platform shows that the proposed method can accurately detect small targets in complex images such as noisehigh-intensity sharp edges and structural backgroundsand its accuracystability and execution time are better than other comparison methods.