Airborne dual-frequency LiDAR detection technology collects simultaneous land and water-depth measurements using dual-wavelength laser. Since 1969, some commercial products have been formed for ocean, coastal zone and reef detection. Since 1998, Shanghai Institute of Optics and Fine Mechanics has successively developed three generations of airborne dual-frequency LiDAR, transforming from principle sample machines to product sample machines. The 11 flights tests of latest Mapper5000 system have been completed in the area of the South China Sea. Topographic and hydrographic data of islands and reefs have been collected. The measured hydrographic range is from 0.25 m to 51 m. The measured hydrographic vertical and horizontal accuracy are 0.23 m and 0.26 m, respectively. The measured hydrographic and topographic spatial density are 1.1 m×1.1 m and 0.25 m×0.25 m, respectively.

Efficiently and quickly measuring shallow waters, islands, reefs, ledges and unreachable waters for ships, an airborne bathymetric LiDAR system is able to measure deep to 50 meters, covering the entire inshore region, while the ship-based mobile measurement system is capable of obtaining topographic data of the ocean floor in inshore shallow sea and fine 3D laser point cloud of islands and reefs. These systems have some complementarity in the area and range of measurement. In this paper, an improved iterative closest point (ICP) algorithm based on feature points of curvature is applied to conduct registration and fusion of airborne bathymetric data, multi-beam data and 3D laser scanning data obtained from the domestic airborne bathymetric LiDAR system and the ship-based mobile measurement system, as well as comprehensively and accurately describe the terrain of both coastal waters and inshore areas for a unified criterion. The results show that the registration data from the airborne bathymetric LiDAR system and the ship-based mobile measurement system can be of great significance in comprehending the topography of islands and reefs, as well as detecting and discovering underwater targets.

As an urban subject, the extraction of buildings has been a hot topic for scholars. Airborne laser scanning data collected in urban areas have a huge data volume and numerous objects with complex and incomplete structures which raise a great challenge for automatic extraction of buildings. To address this challenge, we propose an algorithm based on object-oriented decision tree to extract buildings with high precision. It can handle multiple attributes simultaneously and be unaffected by missed points. First, combined with mean elevation and neighbor within each laser point, judging criteria in each internal node are determined to generate building points by using the relationship between each object attribute and its corresponding eigenvalues. Next, through comparing the entropy from all dataset features, an optimal feature and value candidate are chosen to get a correct classifier with supervised learning to apply to the dataset to be processed. The experimental results show that the proposed method is capable of extracting building points from the airborne laser scanning data with a high accuracy of above 96%.

The effect of crystal material, cavity length, pump power and other parameters on the width of Q-switched pulse is investigated, and a compact dual-wavelength laser with high efficiency based on Nd∶YVO4 crystal is developed. Master oscillator power amplifier (MOPA) structure is used in the laser system. A transverse fundamental mode of 1064 nm pulse with repetition frequency of 5 kHz, pulse energy of 400 μJ and pulse width of 1.1 ns is generated in the oscillator. Through the 878.6 nm zero-line pumped cascade amplifier, pulse energy and pulse duration turn to 2.72 mJ and 1.03 ns, respectively. A 532 nm green laser pulse with energy of 1.54 mJ, second harmonic generation efficiency larger than 56%, pulse duration shorter than 910 ps and peak power of 1.7 MW is obtained through the external frequency doubling by lithium triborate (LBO) crystal.

As light detection and ranging (LiDAR) develops, the extraction of forest structure parameters has been one of hot topics in related fields in the past years. However, the accuracy of detection is the key factor in obtaining the forest individual tree parameters. The individual tree detection methods can be divided to two types: one is based on the canopy height model (CHM) and the other is based on the point cloud distribution. We can identify an individual tree by using the method of the crown boundary segmentation. Also, we can identify the tree top by local maximum algorithms and then perform the regional growth or image segmentation. Based on the point cloud distribution, the canopy is identified by region growing or clustering algorithms in three-dimensional space. We analyze the advantages and disadvantages of different individual tree detection methods in terms of precision of individual tree detection, and compare their effects on omission errors and commission errors in different regions. The factors influencing the precision of data such as data type, point cloud density, season and tree growth status are discussed. It is found that the accuracy of the full-waveform data is higher than that of discrete-echo data. The density of the point cloud data of 10 pt/m2 can meet the individual tree detection requirement. The accuracy of data obtained in winter is higher than that in summer. The limitation of airborne LiDAR data and its shortcomings in individual tree detection are discussed. In the end, the future directions of individual tree detection are described, from the aspects of data acquisition type, data acquisition time, data organization and management, multi-source data fusion, comprehensive application of multi-detection algorithms, and machine learning increasing the training set to find the optimal model, to help with the research and management of forest and related fields.

Aiming at the special oval scanning structure of airborne dual-frequency LiDAR, the error sources that affect the positioning accuracy of the system are mainly studied, including the eccentric errors caused by the installation of the motor rotor nut and the encoder, the installation error of the scanning mirror and optical adjustment error. The error model of airborne dual-frequency LiDAR is put forward. Based on this model, the effects of various errors on the positioning accuracy of airborne dual-frequency LiDAR are quantitatively analyzed. The results show that these errors will have a great impact on the system′s positioning accuracy and have different effects on the position of each point on the same line. After a variety of errors are coupled, it is difficult to set up the system correction model. Therefore, the introduction of errors should be controlled in the design of the LiDAR structure, device selection, structural processing, system installation and other aspects, and the error should be quantified in the assembly process so as to be corrected easily in the later stage. Attenuating and eliminating these systematic errors has an important impact on the improvement of positioning accuracy of airborne dual-frequency LiDAR systems.

Based on the localization model of domestic airborne dual-frequency LiDAR system, various factors that affect the positioning accuracy are analyzed, and the comprehensive accuracy evaluation model of the system is deduced by the error propagation law. According to the sensor parameters of the system, a simulated calculation is made for its theoretical positioning accuracy. The method of airborne radar's location accuracy assessment is analyzed for reference, through which a method to evaluate the accuracy of ground location based on planar features is designed. By using the airfield data and high precision control data to carry out the accuracy evaluation, the thesis gets the result that the accuracy of the domestic airborne dual-frequency LiDAR system meets the design requirement and the evaluation method is simple and feasible.

The existing waveform processing methods in the airborne Lidar bathymetry system are reviewed. Based on the multi-channel advantages of the domestic airborne Lidar bathymetry waveform data and by using the multiple waveform processing methods, such as deconvolution, numerical fitting and signal filtering, a set of waveform processing methods suitable for domestic hardwares is proposed as for different types of waveforms. This set of methods can ensure the robustness of waveform peak position extraction.

A fast atmospheric correction method is used to eliminate the effects of the atmosphere on electromagnetic wave transmission. Aerosol retrieval and atmospheric correction look-up tables are constructed based on the atmospheric radiation transmission model. Aerosol optical depth (AOD) is derived by the dark target algorithm. The atmospheric correction parameters are calculated by the pixels. Using the visible infrared imaging radiometer suite (VIIRS) data on a new polar-orbiting operational environmental satellite, we analyze the difference of vegetation reflectance and water reflectance, and the change of normalized difference vegetation index before and after atmospheric correction. In addition, the results of the fast atmospheric correction are also verified by the measured spectral data and moderate resolution imaging spectroradiometer （MODIS）surface albedo data. The verified results show that the spectral curve of ground object after the atmospheric correction is closer to the measured spectral curve. Compared with MODIS surface reflectance products, the fast atmospheric correction is of high accuracy, and the error is less than ±0.04, so it has more quantitative applications of remote sensing data.

Based on the comprehensive observation data observed in the Beijing southern suburbs observatory, such as aerosol extinction coefficient and particle depolarization ratio obtained from aerosol lidar detection, wind field obtained from Doppler wind lidar, PM10 and ground meteorological elements, we analyse strong sand and dust weather process on May 4-5, 2017. The results show that on the afternoon of May 3, the horizontal wind speed begins to decrease, and the vertical downward airflow appears in the upper air, which are conducive to the dust falling from high altitude to the ground. During 12:00 to 20:00 of May 4, the vertical air flow occurs in the entire height layer from 40 m to 1350 m, and the maximum speed reaches 2 m/s, leading to the spread of low-level dust to high altitude. The three-dimensional distribution of dust is obtained from the lidar product: the initial dust layer height is about 1 km, the mid-term dust height is about 2 km, and the later dust height gets over to be about 1 km. The related coefficient of variation between the extinction coefficient of the lowest observation height of lidar and the variation of ground PM10 concentration is 0.8308. The study shows that the combination of aerosol lidar and wind-measuring lidar can effectively monitor and predict the distribution and diffusion of atmospheric aerosol.

In order to study the aerosol multiple scattering effect on the transmission characteristics of polarized light in the region, we retrieve the microphysical optical parameters of the aerosol, including optical thickness, complex refractive index, particle spectrum distribution, through the inversion of the measured data from the CE318 solar spectrometer. The influence of the complex refractive index of particles, the effective radius of particle swarm and the polarization state of incident light on the transmission characteristics is analyzed through the Monte Carlo vector radiative transfer model. The results show that the transmission characteristics of the smaller particles are more sensitive to the real part change of the complex refractive index. In addition, with the increase of the imaginary part of the complex refractive index, as an important factor affecting the characteristics of radiation transmission, the absorption is stronger and the transmission characteristics is worse. If the effective radius of the particle group is larger, the transmittance of the light wave will be lower and the reflection rate will be higher. The polarization state of the incident light has little influence on the transmission characteristics. Compare to the other types of polarized light, the transmission and reflection rate of horizontally polarized light are greatly affected by the incident angle.

In order to obtain the optical concealment depth of underwater submersible by satellite remote sensing data inversion and realize large scale measurement in time and space, we propose an optical concealment depth remote sensing inversion method based on the quasi analytical algorithm (QAA). At the same time, it is a new way for underwater vehicle equipped with the optical concealment depth measuring device. According to the optical depth concealment model, Aqua-MODIS and Terra-MODIS satellite remote sensing reflectance data is used on the grid. After preprocessing the data, the remote sensing reflectance cross calibration data quality control is completed, and the Terra and Aqua satellite data is cross-corrected. According to quasi analytical algorithm and Doron algorithm, the optical concealment depth remote sensing inversion model is established. The optical depth concealment fusion product for specific sea area is made. Three bands remote sensing reflectance data of 443 nm, 488 nm, and 555 nm is inputted for the model. The data level is L3m, and spatial resolution is 4 km. The longitude range of data is from 100°E to 125°E, and the latitude range is from 10°N to 38°N. The results show that, based on quasi analytical algorithm, optical concealment depth remote sensing inversion technology is a feasible path. A new method to obtain optical depth concealment is developed. The method provides a calibration method for underwater vehicle equipped with the optical concealment depth measuring device, and it also provides important technical support for underwater optical depth detection and anti-detection and other military applications.

The noise characteristics of a low light level imaging system based on the scientific complementary metal-oxide-semiconductor (sCMOS) image sensor are analyzed. The system noise model is established and the system noise is measured. The results show that the root mean square value of system noise is less than 1 e-. The signal-to-noise ratio (SNR) model is built, the calculated SNR of this low light level imaging system is superior to 1 under 10-3 lx night light conditions, and the error between the theoretical and measured results is less than 10%. An adaptive stripe noise processing method is proposed to effectively eliminate the row stripe noise of image under low light conditions and the contrast limited adaptive histogram equalization method is applied to improve the contrast of image with a high dynamic range.

To realize a large scalable optical phased array and beam steering without grating lobes, we design a miniaturized dielectric optical nanoantenna operating at 1550 nm with high efficiency. The nanoantenna is trumpet-shaped with a footprint and composed of silicon in the subwavelength order. Light is fed from the bottom of the nanoantenna by a silicon waveguide and then radiated into free space efficiently. Because silicon is nearly transparent to the wavelength of 1500 nm, absorption loss caused by the material is very low and the nanoantenna has a high radiation efficiency. A three-dimensional electromagnetic simulation software is used for simulation and optimization. The final footprint of the designed nanoantenna vertical to the radiation direction is 0.82 μm×0.28 μm. The results of simulation and optimization indicate that the nanoantenna has good radiation characteristics at 1550 nm with a gain of 6.1 dB, an efficiency of 82.7%, a good symmetry of radiation pattern, emitting vertically to the nanoantenna plane and no bidirectional radiation.

Requests grooming in multi-layer architecture are realized by the joint optimization of IP and optical networks in optical interconnection of data center. Due to the burstiness of requests in intra-data center, burst requests are groomed to optical network, which leads to the longer delay of queue and establishment of light paths of requests. In order to reduce the establishment delay of optical paths, we propose a routing and spectrum assignment (RSA) algorithm with low delay, which separates the spectrum resource of fibers into spectrum layers, and multi threads can set up optical paths in different spectrum layers in parallel. The algorithm sets up colored virtual topology with the points and edges of establishing optical paths in every spectrum layer, and the topology is used as constraints of resource utilization in current spectrum layer. The colored and usable virtual topologies in spectrum layers are reconfigured according to transformations of establishing, unestablished and new arrived requests, which can adapt bursts of requests and promptly groom the requests swarmed in optical network. The simulation results demonstrate that the proposed algorithm can efficiently reduce the average delay, delay variance of establishment of optical paths and blocking probability of requests, and improve the rate of spectrum utilization and quality of experience (QoE).

We have demonstrated a kind of single-end reflective optical fiber Mach-Zehnder (M-Z) interferometer based on large lateral-offset splicing joints. Several hundred microns of standard single-mode fiber (SMF) was spliced between two other SMFs with the same lateral offset. One fiber has its end-face coated by gold film, acting as a reflector. The beams propagating in the environmental medium and the fiber cladding material interfere and produce a phase difference. The reflected interference spectra were measured at the same side of the broadband source by an optical spectrum analyzer. The pattern of interference spectrum shifts with the change of the external refractive index (RI). The sensor with a cavity length of 554 μm and a lateral offset of 62.5 μm is demonstrated. The achieved RI sensitivity in water and the temperature coefficient in air are -13257 nm/RIU and 37.33 pm/℃, respectively. Compared with the reported counterparts, the sensor has the advantages of high refractive index sensitivity, compact structure, single-ended measurement and good stability, and has good application prospects in the fields of biochemical sensing and environmental pollution monitoring.

For identification and decomposition of domestic appliance load, a new feature extraction method is proposed in this paper. First, we adopt wavelet filtering processing on the bus current and remove the background current from the load transient current by periodic difference. Then S-transform is used for the transient current after load switching to obtain an amplitude harmonic matrix. The matrix is dimensionally reduced and the corresponding features are extracted in the row and column directions by the bidirectional two-dimensional principal component analysis. Finally, the feature data is classified by the support vector machine. The experimental results indicate the proposed feature extraction method has an average recognition accuracy of 99.24% and a maximum recognition accuracy of 100% on six kinds of appliance from the BULED dataset. Especially, the proposed method can provide better performance on differentiating the loads with similar electrical characteristics.

Considering the problem that the same frequency interference is easy to occur between the radar signals transmitted by the shipborne radar with same type in the complex battlefield environment, we design a frequency modulation (FM) coded signal with orthogonal characteristics. The search direction of conventional waveform coding with the orthogonal characteristic is improved, and the hybrid genetic chicken swarm optimization (HGCSO) is used to identify the coding sequences with low autocorrelation and low cross correlation characteristics. We adopt the back-learning chicken swarm algorithm to search the optimal values, and introduce the learning factor and the idea of mutation and crossover in genetic algorithm to update the individuals. A comprehensive evaluation of set pair analysis is used in fitness function, and the search direction is guided according to the relational degree of setting pair analysis to get a FM coding sequence pulse signal with excellent orthogonality. The fuzzy function of the obtained radar signal, the matched filtering of the echo signal, and the orthogonality of the simulation signal at different radar numbers are simulated respectively, and the results of simulations show that the FM coded signal designed by this algorithm can effectively resist the same frequency interference.

Based on the recording and reconstruction theory of Fresnel incoherent digital holography, we analyze and calculate the imaging process of the synthetic aperture incoherent digital holography. The effect of mounting the double lens symmetry mode on a spatial light modulator on imaging characteristics of synthetic aperture incoherent digital holography is simulated, and the simulation results are analyzed. We compare two representation methods including different phase sub-holograms which are first spliced and reconstructed or first reconstructed and spliced. The simulation results are verified by experiments. The results show that the central aperture is the most important sub-aperture. The central aperture and four cross-bonding sub-apertures has better imaging performance, or the double lens symmetry mode can greatly reduce the number of sub-apertures and the holographic recording time. The applicability of synthetic aperture technique in incoherent digital holography is verified.

Breast cancer computer-aided diagnosis (CAD) system is playing more and more important role in medical detection and diagnosis. In order to classify tumor and non-tumor in magnetic resonance imaging (MRI), a novel breast cancer CAD system based on deep learning and transfer learning is designed. First, we balance the imbalanced data sets and use data augmentation to deal with it. Then, we use the convolutional neural network (CNN) to extract CNN features from MRI data sets, use the same support vector machine to evaluate the feature extraction abilities of different layers, and select the highest F1 score layer as the node of fine-tuning, the layers behind it, which has relatively low dimension as the node of connection of new networks. Next, we select the newly designed fully-connected layers with two layers to form a new network, and use transfer learning to load weights on the new network. At last, we freeze the layers before the node of fine-tuning, while other layers can be trained in the fine-tuning procedure. The CAD systems are built on three CNN networks, including VGG16, Inception V3, and ResNet50. The effects of the system based on VGG16 and ResNet50 have the best performance, and twice transfer learning can improve the performance of VGG16 network system.

Hyperspectral image have higher and higher spatial and spectral resolution, resulting in a large amount of data, strong correlation and high redundancy, which makes the low accuracy of anomaly detection result. In order to select the image which is more favorable for anomaly detection, we use the two-dimensional principal component analysis (2DPCA) method to reduce the dimension, and introduce the local joint skewness-kurtosis index to image selection. A method based on selective segmentation 2DPCA for hyperspectral image anomaly detection is proposed. Firstly, the original image is segmented by the correlation coefficient, and then the row-column two-dimensional principal component dimension reduction is realized in each band subspace by rotating the data structure. Then, we select an appropriate size window to traverse all the principal components of each dimension reduction result. Meanwhile, the local joint skewness-kurtosis index is calculated in this window, which is regard as an indicator to select the image for anomaly detection. The experimental result shows that the receiver operating characteristic (ROC) curve, the area under the curve (AUC) value and Bhattacharyya distance value of the proposed method are better than other traditional methods, so that it has a better detection performance.

The background difference method based on Kalman filtering cannot adapt to the background update and it is sensitive to light changes and object moving. A modified background subtraction algorithm based on the idea of classification is proposed. First, the initial background model is gotten by averaging the first N frames of the video sequence images. Then, the difference image is obtained from the difference between the Kth image and the background image. The difference image is split into foreground and background blocks for two times and the classification criteria are the mean value and standard deviation. The foreground blocks are finely segmented based on a single pixel, and the moving targets region is determined. Finally, the adaptive background updating is conducted according to the gray value information between adjacent frames. The experimental results show that the proposed algorithm can effectively solve the problems of slow changes in external light and slight movement of objects in the background, and it has good robustness, relatively higher computing speed, and accurate moving targets area.

An adaptive Gaussian attenuation image defogging algorithm based on edge preservation is proposed to solve the problem when transmissivity is estimated under the local similarity assumption, the abrupt change edge of depth field often leads to the Halo effect .This method starts from the atmospheric scattering model and introduces the luminance of the sky, the estimation of scene transmissivity can be equivalent to the estimation of the luminance of the sky. The edge information is extracted by the edge detection operator to separate the edge region and the non-edge region. The adaptive Gaussian function is constructed by using the spatial proximity of the pixel points in the neighborhood to smooth and attenuation on the non-edge region, thereby obtaining the optimal effect. Through a large number of experiments to verify the proposed method, the results show that the restored image is smooth and the details are obvious. It can effectively eliminate the Halo effects at the abrupt changes of the depth field, and it also shows its advantages in the objective evaluation.

The segmentation accuracy of nonparametric semantic segmentation algorithm, is easily affected by the image retrieval accuracy and semantic category unbalanced dataset. To solve these problems, an image semantic segmentation algorithm based on convolutional neural network (CNN) feature and improved superpixel matching is proposed. Image features are obtained through CNN learning, and images are retrieved after reducing dimensions of the features, and then the accuracy of image retrieval set can be improved. Superpixels of the images in the retrieval set are weighted by using Gaussian kernel density estimation, which increases the superpixel matching accuracy of rare classes. Therefore, semantic segmentation accuracy of query image can be improved. The experimental results on SIFTflow and KITTI datasets show that both per-pixel and per-class rates of the proposed algorithm are the best among different algorithms.

In underwater environments, light waves are attenuated by scattering and absorption, which cause problems such as low contrast, blurred underwater image, and color distortion. The color image sensor has three basic color sensors of RGB and each color sensor can detect light information of different wavelengths. Since different wavelengths of light have different transmission coefficients, the differences within these basic colors need to be considered to solve the above problems. Aiming at the problem that the brightness of R channel is decayed as the red band light is seriously absorbed by water, we propose a new method to obtain the distance between the scene and the camera and the attenuation factor of R channel due to water absorption based on electromagnetic wave theory, and ultimately get the transmission coefficient of R channel. In order to validate the feasibility of proposed method, the quality of experimental images is evaluated both by subjective and objective evaluation. The evaluation results show that compared with the traditional algorithm, the proposed method can compensate the image color and remove blur more effectively and make the restored image more realistic.

In order to solve the problem of blurred and lost target edges in the recovery process of fog scene images, we propose a single image dehaze method based on A-Trous wavelet, which is combined with the dark channel prior theory, guided filter, and edge preserving A-Trous wavelet filtering method. The dehaze algorithm of guiding filtering on dark channel prior is reproduced. Then, we introduce edge preserving A-Trous wavelet filter into the dehaze algorithm, and adopt the three level wavelet filtered residuals to compensate for the guide image filtering algorithm with the dark channel prior. Experimental results show that, compared with the guided image filtering haze removal algorithm based on dark channel prior, the rate of new visible edges is increased by 69.5% at least, the normalized gradient mean of the visible edge is increased by 30%, and the percentage of the saturated black or white pixels is reduced by a half at least, when visible edge contrast is used as an objective criterion. The proposed method can retain the edge and texture details of image while removing haze, and the restored image color is closer to the real life. The effect of dehaze is improved, and the edge of the texture is more delicate.

Point registration is a critical step of three-dimensional modeling, but the registration rate has been a major bottleneck restricting development of point registration. In the real life, the point registration data are large in scale and have certain requirement of the registration rate. Concerning decrease of the registration rate resulted from a large point registration scale and potential registration failure caused by a too large cloud distance, this paper combines features of curvature and the concept of coherent point drift to propose a quick point registration method. To begin with, the point cloud curvature is calculated. Then, the curvature similarity between point clouds is compared. The registered point clouds with feature points similar in the structure are extracted. This experiment suggests that this method can not only reduce the time consumption of registration by around two folds, but also register point clouds within the distance of 200 units of coordinate difference.

In view of the most basic sunny days, night and dense fog weather conditions, corresponding filtering of ship images and a scheme for positioning suspicious ships are designed. Gaussian filtering, Laplace image enhancement, and image dehazing are applied to the three different models, which provide better image quality for further ship positioning. The contour feature of the ship is extracted and the number of feature points is reduced by the RDP (Ribosomal Database Project) algorithm. The location of the ship is drawn and calibrated according to the feature extracted. The position of the central point can be extracted. The results show that the gray level of the image is low in the night mode, and the scheme proposed helps to improve the image contrast. The Retinex algorithm provides a feasible scheme for image dehazing. Feature extraction for ship positioning also has a good practical effect.

A dark channel prior dehazing algorithm based on sky-potimization is provided to deal with the problem that dark channel prior dehazing algorithm is limited to the sky region. On the basis of the original formula, the sky region constraint factor is added according to the proportion of the light median value of the sky atmospheric component and the sky region, and makes it suitable for the processing of the sky region. At the same time, the sky region is obtained by K-means clustering and segmentation image, and the smoothed result of the sky region is used as a new guide image to guide and filter the transmittance diagram based on the original haze image as the guidance filtering for transmissivity. Experimental results show that the proposed dehaze algorithm is significant improvements whether in the sky region and the actual dehaze effect, or in the overall visual perception of the image compared with the classic dark channel prior dahazing algorithm and other dark channel prior improved algorithm.

In combination of the perceiving characteristics of human eyes for salient objects and moving targets, a video quality evaluation method based on time-weighted processing is proposed. Firstly, the quality score of single frame is obtained by using the classical structural similarity algorithm, and then the quality score of each group of pictures (GoP) is weighted to obtain the GoP mass score according to the influence of different types of frame distortions on the video quality in the video. Finally, all GoPs are time-weighted with saliency and temporal perceptual information, and the weights of both are equalized to get the final objective video quality evaluation value. The proposed metric is tested on the LIVE video database. The experimental results show that this method has good consistency with the subjective evaluation results, and the evaluation results are better than the existing classical models.

In the process of restoration of ceramic artifacts, the fractured site is easy to be damaged due to its own characteristics, the environment, and man-made factors. The traditional stitching method based on geometric information of fracture location has some limitations. We propose a method of splicing of fractured sites combined with surface texture and boundary contour for the restoration of fragments of ceramic artifacts. Firstly, the potential ridge points are determined by the curvature and the least squares method is used to extract the texture features of the debris surface. Then, the texture elements are combined with the edge length and the vertex angle information to construct damaged texture constraints and realize preliminary matching. Two adjacent points in the boundary set of the contours are concatenated to form the contour chord sequence descriptor, and the contour position information is constrained to obtain the final matching pair. The results show that the proposed algorithm can achieve the splicing of broken artifact fragments, and it has certain advantages for the debris with obvious texture features.

Polarization includes rich information of the reflection and scattering properties of the objects, which can effectively improve the detection capability. Previous studies on the polarization imaging of underwater objects mainly focused on the degree of the polarization, however, the capacities of Stokes components for imaging the underwater objects are poorly known. Based on the tank experiments, the influences of object depth, sediment concentration, object material and detection band on the polarization imaging detection are investigated. The results show that compared with the traditional imaging based on radiation intensity, the polarization imaging can obtain clearer image and better information of boundaries and textures of underwater objects, with the capacity of diminishing the impacts of the water absorption and scattering effects. Moreover, the detection capacity of polarization imaging with increasing object depth deceases more slowly than that of the traditional imaging based on radiation intensity, indicating that polarization imaging can detect deeper objects. The influence of sediment concentration on polarization imaging is more significant than that on the radiation intensity imaging, but the polarization imaging can still detect the edge contour of the target in the high concentration of sediment. The detection wavelength has some influence on the capacity of the polarization imaging, and the blue and green light wavelengths are the optimal bands for polarization imaging of the underwater objects.

The image quality of the surveying camera will affect the stereoscopic positioning accuracy of the satellite remote sensing. The key factors closely related to the image quality are modulation transfer function, signal to noise ratio and quantization bits. Taking Mapping Satellite-1 image as the background, we research the effect of positioning precision about the image quality in no ground controlled conditions, and evaluate the quantitative relationship between the image quality and the stereoscopic position accuracy. At last, the validity of the experimental results is verified by simulating three factors of degradation data in orbit.

In order to meet the engineering requirements of long haul time transfer over optical fibers, a time-delay measurement method based on the time digital conversion (TDC) and the field programmable gate array (FPGA) is proposed. This method combines the characteristics of a large measurement range for FPGA and a high resolution for TDC, and thus an optical fiber time-delay measurement with a high resolution in a wide range is realized. The research results show that the measurement range of this system is 0-1 s, the resolution is 22 ps, and the uncertainty degree is superior to 100 ps.

Position sensitive detector (PSD) has a great advantage in terms of visual pose measurement due to less affect by light and the simple data processing. However, due to the non-linearity of the PSD sensor itself, the distortion of the PSD sensor is complex and the calibration of the PSD sensor becomes the key to its application. Aiming at the PSD vision measurement system composed of PSD camera and infrared light emitting diode (LED) light source, we propose a PSD camera calibration method. The method uses polynomial model to characterize the PSD camera distortion, combined with the traditional pinhole imaging model to form the PSD camera imaging model. Non-linear optimization is used to get calibration camera model parameters. A dedicated infrared LED calibration board is designed to achieve PSD vision measurement system calibration. Experimental results show that this method has high calibration accuracy for PSD camera. With this calibration result, we can obtain high measurement accuracy of position and orientation and short running time, which can meet the demand of high precision optical measurement.

In order to simplify the monocular measurement system calibration tools and quickly adjust the scanning direction of measurement system, we design the “cross” calibration plate and pose-adjustable working stage in the experiment of measuring the cladding layer size through the monocular line structure-light sensor. The “cross” calibration plate can be used to calibrate the camera and the laser plane. The combination of the “cross” calibration plate with the pose-adjustable working stage is not only helpful to calibrate the scanning direction of the measurement system, but also avoids the errors caused by the deviation of the laser projector. In the whole calibration process, only the same calibration plate is required, so the experimental tools are simplified. Moreover, the pose-adjustable working stage satisfies the condition of fast calibration and real-time adjustment of scanning direction. The device is simple to operate, and presents intuitive calibration results. The measurement error of the cladding layer is less than 0.05 mm.

In order to improve the measurement precision of the homodyne laser interferometer for vibration measurement, we propose a real-time active nonlinear compensation method based on the pseudo-extremums. According to this method, the pseudo-extremums instead of the extremums are used to determinate the direct current (DC) offset and alternating current (AC) amplitude difference. Then the quadrature phase shift is suppressed by the operation for vector phase correction. Experimental results show that the method proposed can reduce the error caused by the bit error of digital signal transmission, laser power drift and Abbe error during measurement. Laser power drift and Abbe error make the elliptic trajectory of quadrature signals into a spiral shape. The peak-to-valley amplitude of residual error can be reduced to 0.8 nm by using the pseudo-extremums method with the operation for vector phase correction, which has a more effective suppression for nonlinear error than the conventional extremums method. Besides, the method proposed does not need complex calculation and remains good instantaneity.

The optical system of the head-mounted display is prone to image distortion generally because its focal length is short and its field of view is wide. An image distortion measurement method of head-mounted display is presented. A Fourier transform image of unit matrix is chose as standard test pattern, and the relationship between optical transfer function of head-mounted display and spherical distortion of image is analyzed with Fourier transform. Based on this method, the similar relationship curve are obtained. Image spherical distortion of five kinds of head-mounted display are measureds, and the negative-pinched pincushion distortion is obtained which ranged from -22% to -65%. The results show that spherical distortion CVR of the head-mounted display has linear relation with distortion quality of grid image measurement. However, the measuring method for spherical image distortion is simple, and the measuring results can reflect the distortion of the whole image. The results are of great reference value for measuring image quality of head-mounted display.

To achieve single-mode optical microfluidic laser, we prepare Fabry-Perot(F-P) optical microcavity with high quality factor and use the organic dye Rhodamine 6G (R6G) dissolved in liquid as gain medium. The full width at half maximum of laser output is 0.260 nm. In the following experiments, the generation of single-mode lasers in different mixed solutions (absolute ethanol and deionized water) with water molar fractions of 1.09%，5.98%，11.91%，20.42%，30.75%，45.27%，51.89% is achieved, respectively. It is found that the center wavelength of the single-mode laser moves toward long wavelength as water contents increasing. However, when the water molar fraction exceeds 45.27%, the center wavelength of the single-mode laser starts to move toward the short wavelength. Measurement of mixed solution refractive index is realized by single-mode laser wavelength movement. It is found that with the increase of water content, the measurement and analysis of the refractive index of the mixed solution show that the minimum measurable difference of refraction indexes is 6.31×10-4 and the measurement sensitivity is 411 nm/RIU, in which RIU is refractive index unit. The measured refractive index results and errors are analyzed.

Based on the objective fact that the effect of thermal lens and the influence of spherical aberration coexist, the intrinsic theoretical relationship between the beam quality factor M2，the thermal focal length and spherical aberration is established, and the beam quality change under the combined effect of the thermal focal length and the spherical aberration is more accurately analyzed in addition. In this experiment, the Hartmann-Shack wavefront sensor is used to reconstruct the related wavefront information of the Nd∶YVO4 crystal by LD end-pumped. The experimental data of the beam quality factor M2 are compared with the theoretical analysis. Furthermore, under the influence of the thermal focal length and the spherical aberration, the gradual attenuation of the beam quality is proved. The feasibility of this model is finally confirmed. The above conclusions provide the reference for the optimal design and experiment of solid-state laser.

A short wavelength red light 640 nm high power laser diode has been designed and fabricated. AlGaInP epitaxial layers of the laser diodes are grown by metal organic chemical vapor deposition. The cladding layers are AlInP with low refractive index. The active layer is tensile strained GaInP/AlGaInP quantum well. The photoluminescence spectrum of the active layer shows two splitting peaks locate at 627 nm and 616 nm, which correspond to the transitions from electrons to light holes and heavy holes, respectively. Zn atoms are selectively diffused into the window region, leading to the mixing of the quantum well. The wavelength is blue-shifted by 43 nm. The catastrophic optical damage (COD) occurs for the laser diode without window structure at 1.9 A, corresponding to the power of 1.4 W. The device with window structure has no COD phenomenon. The output power is limited by the thermal rollover with the maximum of 2.3 W. At room temperature, the wavelength of the laser diode is 639 nm at 1 A while 640 nm at 1.5 A. The horizontal divergence angle of the device is 6° and the vertical divergence angle is 41°.

To overcome the disadvantages of traditional Pound-Drever-Hall (PDH) laser frequency stabilization method, we have design a PDH laser frequency stabilization system based on orthogonal demodulation method.The system uses the same direct digital synthesizer (DDS) to synchronously generate three sine signals with the same frequency. One signal is used as local oscillator signal to drive an electro-optic modulator so as to produce the phase sidebands, and the other two sine signals with a phase difference of 90° are used as demodulation references. The in-phase and orthogonal components of the error signal are obtained by use of two analog demodulators, and the error signal of the frequency stabilization system is obtained when both components are digitally sampled and then operated by an algorithm of digital phase-sensitive detection. The key techniques for the PDH laser frequency stabilization based on the orthogonal demodulation method have been investigated, and an experimental system of laser frequency tracking has been established. Experimental results show that the F-P cavity can track the laser frequency variation in real time, with a tracking time of approximately about 1 hour.

Based on the laser additive manufacturing technology, the thin-walled structural parts are formed. The method of negative defocusing of powder is adopted to solve the both-end collapse problem in the forming process of thin-walled structural parts. The microstructure and mechanical properties of thin-walled structural parts are analyzed. The results show that the ideal morphology of single track cladding layer is obtained at laser power of 1400 W, scanning speed of 0.6 m·min-1 and feeding speed of 9.5 g·min-1. The surface of the thin-walled structural part is free of powder and has no oxidation color when the lifting capacity of single layer is 0.57 mm. The height of cladding layer and the using efficiency of powder decrease with the increase of positive and negative powder defocusing. The microstructure of the thin-walled structural part is mainly dendrites growing epitaxially and the dendrites near the substrate are relatively coarse, however at the top is equiaxed grain structure. The hardness of the thin-walled structural part is higher than that of the substrate, and the hardness near the substrate is relatively smaller.

Increasing demands for accurate reconstruction of the non-cooperative target are arising recent years. The existing methods which combine cameras and laser range finders (LRF) are inconvenient and costly. A new vision system is proposed that the widely used laser range finder is replaced with a simple laser. The combination of a camera and a simple calibrated laser is also highly accurately enough to reconstruct the three-dimensional position of the laser spot, while the cost much less than LRF. Also, a method to calibrate the extrinsic parameters between a camera and a simple laser is proposed. First, the direction and the position of the simple laser under the coordinate system of the camera are attained, and the extrinsic parameters of the camera and laser system are gotten using the calibration algorithm. Then how to use it to reconstruct a laser spot′s three-dimensional position is displayed. The experimental results show that the accuracy of the proposed method is comparable to the state-of-the-art LRF-based methods.

In view of the difficulty of full-scale measurement of large workpieces, a portable method of space point positioning based on structure-from-motion (SfM) is proposed. First, the coded target with unique identity is pasted on the surface of the measured object to achieve the stable matching of the corresponding points of each view. Then, to avoid the cumulative error caused by the position transformation in different coordinate systems, a unified coordinate system is selected, and a reference location algorithm based on the center of barycenter constraint is proposed. On this basis, the robust estimation of the camera motion parameters is further determined, and the 3D coordinates of the mark points are reconstructed by the multi view geometric constraints, and the global optimization of the 3D reconstruction results and the intrinsic and extrinsic parameters of the camera is carried out by the beam adjustment. The experimental results show that the method can achieve high precision measurement for large workpieces. The maximum error of 3D measurement is 0.133 mm and the average error is 0.031 mm, which can meet the requirements of industrial field measurement precision for large workpieces.

A series of Li+, Bi3+ co-doped Lu2O3∶Ho3+ phosphors are prepared by the high temperature solid-state method. The research results show that the doping of Li+, Bi3+ and Ho3+ with different mole fractions does not change the cubic phase structure of Lu2O3. Compared with that of Lu2O3∶ 2%Ho3+ sample, the luminescence intensities of the three kinds of samples doped with 16%Li+, 1.5%Bi3+, 2%Li+/1.5%Bi3+ increase by 3.0, 128.9, 1.4 times, respectively. However, the fluorescence lifetimes of these three kinds of samples under the excitation of a laser with a wavelength of 449 nm are shortened by different degrees.

The intravascular ultrasound (IVUS) and the intravascular optical coherence tomography (OCT) both belong to the non-invasive medical imaging techniques, which are often used for the examination of thin cap fibroatheromas (TCFA) in the human coronary arteries. The detection depth of IVUS is large, so the size of TCFA can be measured accurately, but due to the lower resolution, the fiber cap thickness of TCFA cannot be measured, whereas the OCT has high resolution, which enables it to accurately measure the thickness of the TCFA, but it cannot describe the size of TCFA because of small detection depth. By combining IVUS with OCT, ultrasound and OCT tomographic images of the vascular wall can be obtained simultaneously in a single examination, and the accurate diagnosis of intravascular thrombosis and plaque is achieved. In this paper, we design and realize IVUS and OCT integrated micro-scanning probe and near-end driving module. Meanwhile, a set of system that can simultaneously perform IVUS and OCT imaging on the blood vessel wall is realized. The detection depth of the blood vessel tissue reaches 8 mm, the lateral resolution is better than 1 mm, the axial resolution is better than 400 μm, the lateral resolution is better than 10 μm and the axial resolution is better than 8 μm in the 1 mm range. Vascular models are built with agar and wire, and imaging experiments are performed to verify the feasibility of the system.

A primary-secondary collimating illumination design method based on compound parabolic concentrator and freeform lens is proposed in order for using LED light source to achieve uniform illumination within a specific region. This method can confine the exit angle of the system in the range of ±5°-±15° and achieve a uniform illumination in the target region. Simulation results show that when the exit angle of the emit light is between ±5° and ±15°, both of the energy efficiency and the illumination uniformity can exceed 85% under the ideal conditions. Also, the illumination uniformity of the system will not change significantly with the alter of illumination distance, and we can change the illumination distance to change the spot size and spot brightness in practice.

In order to help cataract patients obtain a better visual experience with the intraocular lens implantation, which form a clear image on the retina in the far, intermediate and near vision synchronously, we design a trifocal intraocular lens with freeform-Fresnel surface. Each individual focal length is controlled by several different Fresnel zones, which can avoid the focal point disappearing with pupil contraction. Based on the vector form of refraction law, the Fresnel lens with freeform surface is designed by iterative computation, and the energy distribution at different focal points is simulated. Afterwards, we construct the trifocal intraocular lens with freeform-Fresnel surface in the eye model of ISO 11979-2, of which the additional power of the focal length is +1.66 D and +3.32 D, respectively. By ray tracing, the modulation transfer function of the system at different focal point is obtained. It is clear that the modulation transfer function is greater than 0.2 at the spatial frequency of 50 lp/mm for all focal points, which meets the daily vision demand of cataract patients.

Based on the thermodynamic properties of chalcogenide glass material Ge23Se67Sb10, the molding process model is established. The effects of molding temperature, friction coefficient and molding rate on the equivalent stress of chalcogenide glass after molding are simulated. Through the comparison with the molding simulation data of L-BAL42 material, the difference between chalcogenide glass and traditional visible light glass in the molding process is discussed, and the parameters of the molding process are optimized. The research results show that the equivalent stress decreases with the increase of molding temperature，the decrease of friction coefficient and molding rate. Compared with the control material, the chalcogenide material Ge23Se67Sb10 is the most suitable for changing the equivalent stress by changing the molding rate.

An AlN-based piezoelectric continuous membrane deformable mirror is designed. Based on the theoretical analysis and COMSOL simulation, the structure of AlN MEMS deformable mirror is optimized and the first 14-term Zernike polynomials are fitted. Its driving performances, especially its aberration correction ability, are investigated systematically. The research results show that such a deformable mirror has the advantages of excellent linear control and large deformation. It is compatible with the COMSOL process, and its processing technique is simple and environmentally friendly, which meets the general requirements of self-adaptive optical system.

Aiming at the beam optimization problem of waveguide optical phased array, we propose a sidelobe optimization method with mixed weighting of amplitude and phase based on genetic algorithm. Taking one-dimensional linear array as an example, we analyze the implementation method of mixed weighting. On the basis of the analysis, the MATLAB simulation in computer is carried out for the proposed genetic optimization algorithm. The simulation analysis shows that the optimized antenna pattern compresses the grating-lobe to the sidelobe. Meanwhile, the peak sidelobe level is greatly ameliorated. We perform a comparative analysis of whether the main lobe is included in the optimized search process, the optimized simulation results show that both of them can achieve the purpose of compressing the sidelobe. This algorithm has good convergent performance and can improve the efficiency and quality of the design by optimizing the design of the direction diagram. It has very good practical engineering applications in the direction diagram synthesis.

In order to enhance the light-emitting diode (LED) light extraction efficiency, the influences of grating shapes (rectangle, isosceles trapezoid and isosceles triangle) on the extraction efficiency are analyzed. The LED integrated with different shaped gratings is optimized based on the equivalent medium theory and the rigorous coupled-wave analysis method. The light extraction efficiencies for three LEDs with different optimal structures are simulated by the finite difference time domain method. The research results show that, when the wavelength is within the range of 0.4-0.5 μm, the light extraction efficiency of a flip-chip LED with an isosceles triangle shaped grating is larger than 70%, and the maximum is up to 77.75%, while that of a flip-chip LED with a trapezoid shaped grating is the lowest and its maximum is only 58%.

The actual deflection efficiency of the liquid crystal optical phased array is far below the theoretical level. In order to solve this problem, the bat algorithm is compared with the particle swarm algorithm from two aspects of the number of iterations and the diffraction efficiency, respectively. The simulation results show that, as for the bat algorithm, its optimization effect is fine, and its iteration speed is larger than that of the particle swarm algorithm. The bat algorithm possesses a good application prospect in the beam optimization of phased array.

The extinction property and electric field distribution of the metal-compound-graphene composite nanoarray are simulated. The research results show that the dipole resonance peak intensity gradually increases with the increase of the atomic layer number of graphene. The dipole resonance peak position is blue-shifted from 680 nm to 500 nm when the two laminate structures and the graphene thickness are changed. The refractive index of external media has a significant influence on the resonance peak of LSPR. When the refractive index is fixed, the electric fields at the base angle between graphene column and metal Ag board and at the apex angel between graphene column and SiO2 board are strongly enhanced.

A method of target recognition based on the blur-invariant convolutional neural network (BICNN) model is proposed. The BICNN model introduces a new blur-invariant layer, which is different from the traditional convolutional neural network (CNN )models. BICNN is trained by the adding of the blur-invariant constraint term and the regularization to optimize a blur-invariant objective function. The value of the fuzzy invariant objective function is reduced to make the training samples consistent with the feature maps before and after the blurring, and thus the blur invariance is achieved finally. The test results show that BICNN can solve the problem of a low recognition rate caused by blur and improve the recognition rate of the motion blurred images.

The depolarization theory and numerical calculation model for a dual Babinet depolarizer applied to rectangular pupils are built, which makes up the deficiency of the current design theories only suitable for circular pupils. A type of dual Babinet depolarizer with high reliability which can be applied to rectangular pupils is designed. On the basis of the optimal structural parameters，the error requirements of the refractive index and the central thickness difference of crystals are analyzed. The results show that the maximum residual polarization degree of the output light from the depolarizer is less than 2% and the maximum separation distance for the four images is 27 μm. The design requirements of polarization sensitivity and image quality are satisfied.

Soft glass fibers have been widely applied in generation of mid-infrared supercontinuum, and have become a research focus. This article reviews the research progress of mid-infrared supercontinuum in fluoride fibers, tellurite fibers and chalcogenide glass fibers. The highest power of mid-infrared supercontinuum has been achieved in the fluoride glass fiber; tellurite glass fibers (in particular, microstructured) are widely applied in the mid-infrared supercontinuum; and chalcogenide glass fibers have been found to possess the broadest mid-infrared supercontinuum.

Narrow linewidth fiber lasers have gained extensive attention as their important applications in geoscience, nonlinear frequency conversion, and beam combining. Unfortunately, the output power of narrow linewidth fiber lasers is limited by nonlinear effects such as stimulated Brillouin scattering (SBS), self-phase modulation (SPM), and four wave mixing (FWM). Lots of methods have been proposed to suppress the nonlinear effects, which boost the output power of narrow linewidth fiber lasers up to kW level. In this paper, we give an overview of the research on high power narrow linewidth fiber lasers at the wavelength of 1 μm. The nonlinear effects and corresponding suppressing methods are introduced, as well as the recent progress and key factors of high power narrow linewidth fiber lasers.

Chalcogenide glasses have ultra-high nonlinear polarizability and ultra-short nonlinear response time. Because of excellent thermal stability, chemical stability, and fiber forming properties, chalcogenide glasses have attracted extensive attention in the field of infrared photonics. In this paper, the latest research progress of the third-order optical nonlinear properties of chalcogenide glasses is reviewed in terms of glass composition, microcrystalline treatment, and light irradiation. The current research status of the third-order optical nonlinear properties of chalcogenide glasses is summarized and prospected.

Dioxin is the most toxic chemicals known at present, it is usually generated in combustion and industrial production processes, and it is difficult to degrade, therefore, it′s particularly important for its detection. Infrared spectroscopy can retrieve component concentration based on infrared absorption characteristics of molecules, which has been widely researched and applied in small molecules. At present, there is no information about the parameters of dioxin infrared spectrum, in order to measure the concentration of dioxin by spectroscopy, firstly, we need to study the spectrum of dioxin. The spectral calculation of 2, 3, 7, 8-tetrachlorodibenzo-p-dioxin (TCDD) molecule is carried out with Gaussian 09 software package. The VERTEX 80 Fourier transform infrared spectrometer is used to measure the 2, 3, 7, 8-TCDD molecule spectrum in the range of 1600-1000 cm-1 under different concentrations of 30 μg/mL, 40 μg/mL, and 50 g/mL, respectively. The resolution of the instrument is 4 cm-1 and the scanning times is 16. In addition, the spectral intensity of 2, 3, 7, 8-TCDD molecule at 1466 cm-1 is measured at 297 K. The results show that the maximum absorption peak of 2, 3, 7, 8-TCDD molecule is located near 1466 cm-1, this is caused by the vibration of benzene ring. It has good correspondence of spectra obtained by theoretical calculation and experiment. This study provides spectral information for the measurement of 2, 3, 7, 8-TCDD molecule concentration by infrared absorption spectrometry, and has important theoretical significance.