In order to realize the lightweight of the (165mm×96mm) rectangular scanning mirror assembly and ensure the accuracy of the mirror shape and the support stiffness of the assembly, a back support method with the integration of cone-sleeve and flexible joints was proposed, and the weight of the mirror assembly designed with this method is less than 0.5kg. The mirror body was made of silicon carbide and the supporting structure was made of indium steel. The support stiffness and surface accuracy of the scanning mirror assembly are simulated and analyzed by finite element analysis and verified by actual testing. The results show that the maximum root mean square (RMS) value of the scanning mirror shape error is 9.705nm, the actual test result 10.125nm, the error is 4% and meets the requirement that the RMS value should be less than 12.6nm in the case of gravity in all directions and the torque of the shaft-driven. The first-order natural frequency of the component was 302.25Hz, which met the stiffness requirements. The results show that the proposed method is reasonable and effective, which solves the problem that the ultra-lightweight structure and the structural stiffness and optical surface accuracy are difficult to guarantee at one time.

Optical fluorescence microscopy has gradually become a key device in the field of biomedicine, due to its low phototoxicity and easy 3D imaging. This system is a large-value aperture microscope objective designed by ZEMAX, which is characterized by imaging optimization for commonly used fluorescence wavelengths (460nm/509nm/590nm/620nm), and introduces a diaphragm structure to make the focal depth of the objective adjustable, thereby improving the adaptability to the thickness of the optical chip, reducing stray light interference, and increasing the signal-to-noise ratio of the system. The microlens array is designed at the end of the system, which refers to the pixel size adjustment of the CMOS sensor, which improves the energy utilization of the system. The total length of the objective is less than 45mm, the numerical aperture is 0.7, the magnification is 40x, and the working distance is 0.4mm, all of which meet the national standards.

One of the most important safeguards for flight safety is the detection of foreign object debris. In view of the development trend, the multi-source detection system of "radar + photoelectric" is the development direction of foreign body detection on airport runway. It is urgently needed to have a high integration and coordination of integrated equipment. Based on the principles of photoelectric detection and millimeter wave detection, designed through irregular antennas, the detection capability of millimeter wave and optical detection is efficiently allocated, and the configuration of millimeter wave antenna aperture and optical detection lens is reasonably planned, the co-design and optimization of optical imaging system and millimeter wave antenna are completed, and the integrated detection antenna of millimeter wave and photoelectric composite is obtained. This composite antenna ensures the maximum detection capability of optical system and millimeter-wave radar and obtains a compact and reasonable integrated structure, which lays a technical foundation for high integration and high performance multi-source detection system.

The measuring range of most existing dispersive objective lenses for the chromatic confocal microscopy are on the order of several millimeters. In order to meet the demand for large range linear dispersive objective in the industrial measurement field, a linear dispersive objective with a super large range was designed based on the design principle of linear dispersive objective. It is composed of only five spherical lenses, all of which are made of cost-effective and environmentally friendly optical glasses. The design results show that the axial dispersion of the objective lens is 30.44mm in the wavelength range of 400~700nm, and the dispersion-wavelength linearity is higher than 0.99, which has an excellent linear relationship, and the theoretical resolution can reach 2.034μm.

Fused quartz glass is widely used in aerospace, microoptical components, military and other fields because of its high heat resistance, low thermal expansion coefficient, good insulation properties, and so on. Higher requirements are put forward for its machining precision and surface quality. Because femtosecond laser has the characteristic of "cold processing", it has shown unique advantages in the micro-nano processing of fused quartz glass. Fused quartz glass was processed by femtosecond laser with wavelength of 1030nm, repetition frequency of 100kHz and pulse width of 290fs. The damage threshold of fused quartz glass under different objective lenses was determined. The effects of laser power, scanning speed, defocus and scanning times under different objective lenses on the machining line groove were studied. With the method of layer-by-layer superposition machining, the groove with the ratio of height to width (4∶1) was obtained at low power, and the width and depth of the groove were improved, and the micro-nano machining could be carried out on the thinner fused quartz glass (200μm).

In order to solve the problems of time-consuming processes and insufficient accuracy in the surface porosity recognition technology of the cladding layer, A 2BNC-Unet neural network based on the U-Net neural network is proposed. The goal is to identify pores on the cladding layer's surface using semantic segmentation in deep learning technology. By introducing the Batch Normalization layer and the Convolutional Block Attention Module (CBAM) into the neural network in a reasonable manner, the Intersection over Union (IoU) and Dice coefficient were selected as evaluation indicators for the network. The results show that, in the test set, the intersection over union and Dice coefficient of the 2BNC-Unet network are 86.96% and 86.42%, respectively, which are 7.65% and 4.73% higher than those of the U-Net neural network. Additionally, to verify the performance of the network, comparative experiments were conducted using SegNet, 2BNC-Unet, and U-Net neural networks. The results demonstrate that the segmentation effect of 2BNC-Unet is not only better than that of SegNet and U-Net networks but also capable of completely segmenting the pore details on the cladding layer's surface. In the realm of deep learning technology, the segmentation speed and accuracy of 2BNC-Unet have been significantly improved, providing assistance in the performance analysis of cladding layers through pore segmentation.

In order to solve the problem that the pulsar recognition research is limited to the field of conventional classification algorithm, the research ideas are mainly divided into two categories: data adaptation model and model adaptation data, and the lack of pertinence is the problem, to study the characteristics of the pulsar dataset, excavates the correlation between its intrinsic characteristics and other research fields, and finds that there is a connection between pulsar data and long-tail distribution. Aiming at the limitations of the field of pulsar data identification, combined with the characteristics of pulsar data itself, in this study. The consistency of pulsar data with the long-tail distribution is explored, and for the first time, the pulsar data distribution is regarded as a special case of long-tail distribution. On this basis, from the perspective of optimized training strategy from the perspective of long-tail visual recognition, a pulsar recognition algorithm based on decoupling training strategy is proposed. The traditional algorithm in the field of pulsar recognition is mainly improved from the idea of optimizing the model and data, compared with the traditional algorithm, the algorithm proposed starts from the training point of view, adopts the decoupling training strategy, and is simple and efficient in operation, with stronger portability. The training process of the algorithm is divided into two stages, the first stage is the joint training of the overall pulsar recognition model, and the sampling strategy is sampling based on the balance of the instances; The second stage is to fix the feature extraction network on the basis of the first stage, and fine-tune the classifier. A variety of different fine-tuning strategies were used, including class balancing training on classifiers and normalization of features to find nearest neighbors. After the verification of multiple data sets, the individual fine-tuning of the classifier can effectively improve the decision-making boundary and improve the recall rate and other indicators, which is a cost-effective improvement method.

To investigate the effect of water stress on chlorophyll fluorescence parameters of tea seedlings, Longjing tea seedlings were selected for water stress experiments. The slope fluorescence index SFI was proposed to assess the drought stress condition of tea seedlings by using chlorophyll fluorescence imaging technique, and the polar difference R of the mean gray value at the same time within 1~8 days was calculated, and the relative error of R was used to determine the stable time range of the polar difference, and the correlation model between the change of slope of each point on the curve and the number of days of stress was established in this interval. The correlations between the slope fluorescence index SFI, dynamic fluorescence index DFI, photosynthetic function index PFI, and fluorescence decay ratio Rfd and the number of days of stress were analyzed and compared. The results showed that there were no significant correlation between Rfd, PFI and days of stress; and there were correlation between SFI, DFI and days of stress. The correlation between SFI and days of drought stress (coefficient of determination R2=0.94) was significantly higher than that between DFI and days of stress(R2=0.60) when the ROI of the region of interest was 50×50pixel. The use of SFI to assess the drought stress status of tea tree seedlings has higher accuracy and provides a reference for monitoring plant growth in modern agriculture.

In order to accurately and quickly measure the axial gap between each lens in the mirror group, a differential confocal gap measurement method based on five-dimensional automatic adjustment is proposed. This method detects the translation deviation and tilt deviation of the optical axis of the lens group under test according to the spot signal obtained in real time by the CCD detector, realizes the high-precision adjustment of the mirror group according to the deviation information, and realizes the high-precision measurement of the axial gap by using the differential confocal high-precision tomography fixed focus and the internal light tracing of the mirror group. Analysis and experimental results show that the method can effectively improve the attitude adjustment efficiency by 3.4 times, and the accuracy of lens gap repeat measurement can reach 0.53μm. This method provides an effective way to achieve fast and high-precision measurement of lens group gap, and also provides an effective method for rapid and high-precision measurement of various parameters such as center thickness, focal length, and radius of the lens.

Aiming at the problem of three-dimensional shape reconstruction of multi-core optical fiber, a three-dimensional shape reconstruction method based on helical multi-core optical fiber and Bishop frame is proposed. The principle of 3D position reconstruction method based on HMCF and Bishop frame was analyzed. The Fiber Bragg grating (FBG) array sensing was used to calculate the strain of each fiber core through spectral drift. The core curvature and torsion angle solution were calculated according to the node section strain relationship model. Four shape reconstruction experiments were carried out, and the maximum absolute error of 3D reconstruction was 3.11mm. The experimental results show that the shape reconstruction method based on Bishop-HMCF can realize 3D shape reconstruction, which has important research significance and application value in the end positioning and navigation of flexible mechanism.

Given the task of quantifying the PV value of the plane shape of an optical element under atmospheric turbulence. Initially, a model of the Fizeau interferometer under atmospheric turbulence is developed, through which 1,000 interference fringe images are obtained. What’s more, a method of estimating PV value based on convolutional neural network is proposed, leting the interference fringes image as the input of convolutional neural network and extracting the feature information from the image to obtain PV value. Finally, the obtained results are compared with those obtained by ASTM(American Society of Testing Materials) calculation, phase unwrapping and BP(Back Propagation) neural network, and the deviation using the method of convolutional neural network is 2.25×10-4λ, which is more smaller than the results obtained by ASTM,phase unwrapping and BP neural network. The experimental results show that the method has strong interference resistance, high precision and fast operation, and is an effective optical detection method against atmospheric turbulence.

In the detection of pavement cracks, the traditional manual detection method is low in accuracy and efficiency, and the monocular vision technology can not accurately obtain the depth information, only two-dimensional plane information. Based on this, binocular line structured light technology is introduced in this paper, and a 3D crack extraction algorithm based on the average ground height is proposed to realize the 3D crack information acquisition. The algorithm extracts the lowest point of each column of pixel values as the starting point, diffuses to the left and right to form an edge extraction range, calculates the average value of pixels outside the extraction range of the column edge as the ground average height value, and then starts to traverse each pixel value with the lowest point as the center. When the left and right pixels exceed the ground average height value, two pixels are determined as crack edges. All columns are integrated to achieve the extraction of the entire crack edge. The experimental results show that the proposed crack identification algorithm can extract the crack edge more accurately. Compared with the Canny edge extraction algorithm, the data show that the algorithm in this paper performs better in continuity. The horizontal correlation data of the algorithm in this paper are generally higher than that of the Canny edge extraction algorithm, with an average increase of 30%.

In recent years, more and more attention has been paid to the detection technology of optical element surface defects. The surface quality of optical element will directly affect the performance of optical system. The detection of micro-defects on the surface of spherical and cylindrical optical elements in curved optical elements are studied, and a detection method based on optical polarization characteristics is proposed. The contrast of defects in the whole image is improved by using the difference of the polarization state of transmitted light between the defect area and the non-defect area. Firstly, based on the polarization theory of light, polarizer is used to obtain polarized illumination light, and confocal illumination is used to obtain the defect image of curved optical element focusing at the same time. Then, using computer to process the defective image. The results show that high contrast and high resolution defect features can be obtained by using polarization characteristics of light to detect micro-defects on the surface of curved optical elements. This method is very good to improve the accuracy and detection efficiency of micro-defects on the surface of curved optical components, and the results show that the detection accuracy of defects reaches 95%.

Usual optical phase retardance measurement methods use a single wavelength of light for measurement, which cannot accurately reflect the variation of retardance amount with wavelength. A simple method for measuring phase retardance spectrum is proposed. The method uses plane polarimeter to measure the transmission spectrum of the system by rotating the component to be measured, and then solves the phase retardance spectrum according to the principle algorithm. The experimental results show that this method has good repeatability and simple operation, and the repeatability of the measurement results can be kept within 0.41°.

Micro-pulse Lidar (MPL) technology is an important means of atmospheric aerosol observation, laser-induced fluorescence signals can be used to detect organic aerosols in the environment when using ultraviolet laser light sources. A simulation model for the horizontal detection of organic aerosols by micro-pulse fluorescence lidar (MPFL) is established, and the number of echo photons and signal-to-noise ratio are numerically calculated. According to the simulation results, a micro-pulse fluorescence lidar was designed and built, and the intensity error of the near-field fluorescence echo signal was reduced by calibrating the geometric overlap factor of the system. The test experiment of the lidar system was carried out using the nutritional broth solution as the aerosol sample, and the experiments showed that the spatial resolution of the MPFL system was 7.5m, and the maximum detection distance of the experiment reached 200m. At the same time, a comparative experiment was carried out with another fluorescence lidar with low repetition frequency and high pulse energy, and the comparison results showed that the trend of fluorescence signal intensity received by the two types of lidar had good consistency, and the correlation coefficient was more than 82%. Under the same accumulation time, the standard error of the MPFL fluorescence signal change rate matrix was less than 0.02%, which had better anti-interference performance, and could achieve accurate detection of organic aerosols, which verified the effectiveness and practicality of the system.

The mutual information based visual servo uses the global information of the image to realize servo control, which avoids the extraction, matching and tracking of traditional geometric features, and is robust to light changes and partial occlusion. However, due to the high nonlinearity of its cost function, this method has the disadvantage of small convergence region. A mutual information visual servo method based on principal component analysis (PCA) for dimension-reduced image reconstruction is proposed, which effectively decreases the nonlinearity of the cost function. The convergence region size and convergence speed of mutual information visual servo based on PCA reconstructed image and original image were compared through experiments, and the influence of the number of principal components in PCA is considered. The results show that the mutual information visual servo based on PCA reconstruction image effectively expands the convergence region of mutual information visual servo and has faster convergence speed.

Aiming at the problems of incomplete recognition of forest land boundary area and low accuracy of small forest land segmentation in remote sensing image segmentation by ordinary convolutional neural network, an improved method of DeeplabV3+ network based on transformer and attention mechanism is passively proposed. First, the transformer mechanism is introduced in the encoding stage, and the hole convolution operation in the original pooling pyramid is replaced by a transformer operation that can obtain more context information, thereby improving the network's ability to extract forest boundary information; then, the attention mechanism is introduced. Go to the decoding part of the network to improve the model's ability to detect small forests. Experiments show that the average intersection-over-union ratio (MIou) of the improved method can reach 81.83%, which is 1.25% higher than the original DeepLabV3+ network model. The method fully considers the extraction of forest edge information and the attention to small targets in satellite remote sensing image segmentation, and the improved method proposed can improve the accuracy of forest land extraction in remote sensing images.

In view of the fact that the existing infrared image enhancement methods suffer from under enhancement, over enhancement and noise amplification, an adaptive weighted histogram equalization method for infrared image enhancement is proposed. In the proposed method, the adaptive weighting coefficient of histogram is put forward, which is inversely proportional to the square root of the histogram frequency corresponding to each gray level, appropriately increases the histogram with smaller frequency, and appropriately suppresses the histogram with larger frequency. After the histogram frequency of each gray level is weighted processed with adaptive weighting coefficient, the weighted histogram is equalized, and finally, the enhanced image is achieved by gray level mapping. Experimental results show that compared with existing methods, the enhanced image by this method has higher contrast, is more rich in edge and detail information, its entropy, average gradient and standard deviation are 0.23, 2.2 and 3.7 higher than those of existing methods, respectively. Therefore, this method outperforms the latest proposed existing methods in image enhancement effect.

In order to enhance the retrieval accuracy of person re-identification, a soft mask foreground segmentation and multi information fusion re-ranking based person re-identification algorithm is proposed. A foreground segmentation algorithm based on U-Net is employed to remove the background information of the image while mitigating the sharp boundary transitions in the image segmentation to preserve the key information of the person in the image. In addition, a multi-information fusion re-ranking algorithm based on the siamese deep network is utilized to improve the quality of the retrieval results. Experimental results demonstrate that the two proposed methods are effective supplements to the person re-identification process and can be applied as effective methods to improve accuracy in most existing person re-identification algorithms.