Based on a brief introduction to the current methods of stray light testing and correction, the spectral stray light matrix calibration, which is the core content of the spectral stray light matrix correction method, is introduced in detail in this paper. Then, to calibrate the stray light matrix of the prototype of the forward limb imaging spectrometer, we set up a stray light matrix measuring device in the laboratory. Finally, the accuracy of the stray light correction method based on the stray light matrix of the instrument is verified in an external field experiment.

A detection algorithm based on support vector machine (SVM) is proposed for four-level pulse amplitude modulation (PAM-4) signals in free-space laser communication. Using the power spectral density inversion method, we simulate the atmospheric phase screen. Under the assumption of Taylor frozen flow, we obtain the optical field signal of the laser carrier propagating through the atmospheric turbulent channel. After Gaussian white noises to the signal are added at the receiving end, the PAM-4 signal is judged by the SVM detection algorithm. SVM detection algorithm is independent on the atmospheric channel. Firstly, SVM detection algorithm groups the received signal data, trains the learning and adjusts the parameters through cross validation of each packet data. Secondly, the optimal hyper-planes between every two adjacent levels can be determined by the cascaded two-level classifications. Lastly, the signal judgment can be made by SVM. The bit error rate (BER) obtained by SVM detection algorithm is better than that of the double-step blind detection, which is comparable to that of the optimal bound detection for weak atmospheric turbulences. It is shown that the SVM detection algorithm has a good performance.

Aiming at the problem that the existing surface plasmon refractive index sensor has a small vertical detection depth and the detection range cannot cover the entire thickness of cells, we propose a real-time measurement method of the living cell refractive index for its advantages: large detection depth and high sensitivity. And this method is used to carry out the experimental research on drug susceptibility. Based on the polarization-selective absorption effect, we design and build a graphene-based refractive index sensing system under the condition of total internal reflection. The refractive index with various mass fraction of sodium chloride solution is measured. The results indicate that the sensitivity and resolution of the system is 9.5×106 mV/RIU and 5.5×10-7 RIU, respectively. The experimental study on the drug susceptibility of living cells is carried out by the system. The real-time changes of cell refractive index during the biological evolution of cisplatin and paclitaxel in Ramos cells and Jeko-1 cells are studied, and the consistency of refractive index changes with its pharmacological mechanism is verified.

The application effect of hyperspectral imaging relies heavily on the acquired signal-to-noise ratio (SNR) of images. The hyperspectral imaging requires a high frame rate and a low SNR for a high spatial resolution. The time-delay-integration (TDI) mode cannot be used to solve the problem of weak light energy, because the spectral imaging contains the two-dimensional spatial-spectral information. A swing mirror is usually used to reduce the application requirement. However, not only the volume and weight increase, but also the acquired image is discontinuous and the space reliability of the moving parts is reduced. Thus the ultra-high speed electron multiplication and the imaging spectroscopy are combined organically, and a high-resolution hyperspectral imaging chain model based on electron multiplication is built. This model can be used for a complete analysis of the SNR in an imaging chain, if all elements such as radiation source, object reflectance, atmospheric radiation transmission, optical system imaging, spectroscopic characteristics, detection spectral response and camera noise are comprehensively considered. The LOWTRAN 7 software is used to investigate the atmospheric radiation transmission. The illuminance at the image plane is calculated for different solar altitudes and object reflectance, and the SNR under different working conditions is calculated according to the noise model of an electron multiplying charge coupled device (EMCCD) detector. Through the analysis and experiment of SNR, a suitable choice of multiplication gain can make the SNR of weak spectral signals enhanced by 6 times.

Based on the Thue-Morse aperiodic mathematical sequence, a new kind of diffractive optical elements (DOEs), which is Thue-Morse multi-focal vortex lens, is proposed. Compared with the Fresnel vortex lens, the generated beam by the Thue-Morse vortex lens possesses the optical vortices embedded at the axial foci and the symmetry optical vortices have the equal intensity. The simulations and experiments verify the unique diffraction properties of the Thue-Morse vortex lens, which can generate vortices array and realize the trappings of metal particles and dielectric particles of low refractive index. The Thue-Morse vortex lens can be potentially applied in the research fields of micro and nanophotonics.

The interference between adjacent diffraction orders is unavoidable in the two-dimensional spectrogram of an echelle grating spectrometer in the long wavelength band due to the non-uniform dispersion of prisms. To avoid this drawback and simultaneously take full advantage of the detector's imaging plane, we design a small-size echelle grating spectrometer with a divided spectral coverage is designed The relationship between the detector and the parameters of the above two is established via the detailed analysis of the dispersion principle of echelle grating and prism. The design approach of the echelle spectrometer with a divided spectral coverage is proposed with the design idea of double-slit interval and by the way of double-slit switching. With this design approach, the wavelength range of 165-800 nm for the system is divided into 165-230 nm and 210-800 nm, the focus length is set as 200 mm, and the two-dimensional spectrograms at the two bands are collected, respectively. The ray tracing simulation to the optical system is carried out and the results show that the true spectrum resolution at 200 nm is 0.015 nm, which meets the design requirements.

The sodium-aluminium borosilicate glass and quantum-dot (QD)-doped fiber are fabricated by the two-step high temperature annealing method. The QDs within the fiber have a size of 4.73 nm±0.25 nm and their absorption and emission peaks are at 1450 nm and 1500 nm, respectively. The absorption spectra, photoluminescence (PL) spectra and PL peak intensity versus pumping power are measured. The attenuation coefficient within the fiber at 980 nm pumping power and PL peak intensity and peak wavelength versus fiber length are also measured. The wavelength-dependent attenuation coefficient, pumping excitation threshold and saturation power in the QD-doped fiber are determined. The experimental phenomena is explained from the aspects of energy-level transition, surface effects, and so on.

A PbSe quantum-dot-doped fiber amplifier (QDFA) in NIR region is realized in the experiment. We fabricate the PbSe quantum dot fiber (QDF) with the central diameter of 4.08-5.88 nm by optimizing the heat treatment conditions of the melt-annealing method in the sodium-aluminum-borosilicate silicate glass. The QDFA is composed of the QDF, the wavelength division multiplexer, the isolator, and the pump laser. The experimental results show that the signal light is amplified in the gain wavelength range of 1260-1380 nm depending on the particle size of quantum dots, where signal gain is 16.4 dB and -3 dB bandwidth reaches to 80 nm for the input signal of -17 dBm. An obvious excitation threshold and gain saturation phenomenon are observed in the experiment. Compared with conventional erbium-doped fiber amplifiers and few-mode erbium-doped fiber amplifiers, the proposed QDFA has a lower excitation threshold, wider bandwidth, and lower noise. It provides a novel approach to extend the current optical-fiber-communication waveband and industrial applications.

This study presents a novel methodology to extract feature lines from unorganized point clouds. In this study, the extraction of feature lines from point clouds is divided into two stages: region segmentation and feature detection. In the region segmentation stage, the social particle swarm optimization fuzzy C-means clustering algorithm is introduced to cluster the point cloud data; further, each partition is obtained with a clear boundary, which is beneficial to extract the boundary features. In the feature detection stage, local surface reconstruction that is based on the radial basis function is conducted for each partition. Additionally, the curvature values of the sampling points are calculated according to the established local implicit surface; further, local feature weights that are based on the mean curvature are proposed. The feature points can be identified based on the local feature weights using the curvature extremum method. Finally, the feature lines can be generated by establishing the minimum spanning tree of the feature points. Different point cloud models are selected to perform the feature line extraction experiments. The experimental results exhibit that the proposed method can extract significant and sharp features from the point cloud models along with the curve features with intensity variations.

Among the existing synthetic-aperture radar (SAR) satellites, the GF-3 offers the most kinds of imaging modes. The fusion of the GF-3 SAR images with the multi-spectral images can improve the visual quality of the SAR images. We show how to use the nonsubsampled contourlet transform (NSCT) for simulating high-resolution images such that both the details of the SAR image and the spectral information of the multi-spectral image can be retained. This method ensures that the fusion of SAR and multi-spectral images is not limited by a specific algorithm. To verify the effectiveness of the proposed idea, two types of resolutions are used as the experimental data: the GF-3 satellite SAR images with resolutions of 3 m and 5 m, respectively, and the GF-1 satellite multi-spectral images with a resolution of 16 m. We perform comparative experiments with different fusion algorithms. The results show the effectiveness of the proposed approach. The traditional method that directly fuses the SAR and multi-spectral images can keep the details of the SAR image. However, the noise is obvious and some information of the multi-spectral image remains. The NSCT average images and the average NSCT images can retain the spectral information. The spectral information of NSCT average images is closer to the multi-spectral images than the average NSCT images.

Light is attenuated in underwater environments owing to the scattering and absorption effects. Because of the changes in water turbidity and different depths of the fields in underwater photography, the level of fuzziness and color deviation of the images captured underwater is different. Traditional defogging algorithms appear to have limited effectiveness in the case of these images with varying degrees of fuzziness and color deviation. Therefore, the color compensation based on the bright channel and the image fusion method for underwater image enhancement is proposed to resolve this problem. First, in order to obtain a color-compensated version of the original image, the color compensation based on the bright channel is performed on the original image. This color-compensated image is then subjected to adaptive contrast stretching to obtain a clear image with high contrast. Finally, the multiscale fusion strategy is adopted to fuse the color-corrected and contrast-stretched images. Experimental result shows that the proposed algorithm can be employed in a wide range of applications dealing with multiple underwater degraded images. Furthermore, the proposed method can effectively improve underwater image contrast and balance image color without any prior information.

Underwater photoelectric images have a low signal-to-noise ratio and poor contrast because water absorbs and scatters light. This makes it difficult to identify targets and limits the practical applications and the development of underwater optoelectronic imaging equipment. To improve the detection accuracy and recognition rate of the target, we propose a deep convolutional neural network with one-dimensional parallel convolution and sub-pixel convolution. The convolutional neural network is used to learn the parameters that can improve the image quality from the underwater photoelectric image training set. Then, it can denoise and enhance the contrast for the test images. The peak signal-to-noise ratio obtained using our method showed an average improvement of 2.93 dB over the ratio obtained using the classic denoising and contrast enhancement methods; the root mean square contrast also increased by an average of 14.41. Therefore, our proposed method can effectively balance the denoising, contrast enhancement, and brightness enhancement. This will improve the image quality. The average processing speed of a single image is 9.46 times greater than that of the classic method. Finally, the network is tested using the test set. And our network could improve the image quality and provide a generalization characteristic within a certain range.

A novel image inpainting forensics algorithm based on the deep neural network is proposed, in which the vestigial features can be automatically extracted by the encoder network, the category of each pixel is predicted by the decoder network, and thus whether or not the image is with inpainting and falsification as well as the inpainted and falsified regions can be distinguished. Simultaneously, the feature pyramid network (FPN) is used to supplement the feature map in the decoder network. The MIT Place dataset is used as the training set and the UCID dataset as the testing set. In addition, the different inpainting and falsification algorithms are adopted for the training set and the testing set, respectively. The experimental results show that, compared with the other inpainting forensics algorithms of images, the proposed algorithm has a more accurate inpainting area and a faster processing speed. Moreover, it has relatively good robustness and strong generalization ability against different inpainting forensics algorithms.

Medium-resolution spectral imager (MERSI) obtains global image and radiation data from the second generation polar-orbiting meteorological satellite, FY-3A, launched by China. However, the on-orbit gain levels of imager in shortwave infrared bands occasionally jump due to the unpredictability of space environment and other factors. Such phenomena limit the quantitative application of the data. The space view is selected to identify gain jumps based on the analysis of MERSI observation targets, such as space view, visible onboard calibrator, and blackbody. Automatic identification, gain classification, and level normalization are proposed as methods. The daily identification and intraday precision detection methods are combined to obtain the time and space of gain jumps, and the classified statistical method contraposing the data of gain jumps is used to obtain the on-orbit gain levels. In addition, the onboard calibration source targets and Earth view images are used to verify the normalized effect. The 91 gain jumps of 1.64 μm and 18 gain jumps of 2.13 μm are accurately positioned during the lifetime. In addition, the eight on-orbit gain levels and the levels of all gain jumps are obtained. The results revealed that the normalized effect is ideal. Further, the most gain jumps occur during the Earth view scanning process and the gain jump time varied between the detectors. Because of the complexity of the Earth view targets, it is difficult to realize the normalization of the image of the jumping frame based on the physical methods. These results contribute to the improvement of FY-3A MERSI image quality and play an important role in reprocessing historical data.

The spectral computed tomography (CT) based on the photon-counting detector has a great potential in material discrimination for its ability to obtain the energy spectral information at multiple energy bands. Due to the poor consistency between the narrow energy-bin detection and the photon-counting detector, there are lots of noises and artifacts in the multi-spectral CT images, which is not beneficial to material decomposition and discrimination. Thus, from the point of view of reconstruction, the traditional study method based on tensor dictionary learning (TDL) is improved and a new image reconstruction method based on image total variation (TV) and TDL is developed, which is called TV+TDL for short. This method not only inherits the advantage of the TDL method in enforcing the similarity among all energy channel images, but also further recovers the slim structures and details, effectively suppresses noises, and thus improves the accuracy of material decomposition by introducing the image TV as a regularization term. The simulation results show that the TV+TDL method can effectively reconstruct high-quality multi-spectral CT images and successfully realize material decomposition and discrimination based on the base material model. The validity and practicability of this method are verified.

The retinal vessel segmentation in color fundus images is of great value for the clinical diagnosis and a retinal vessel segmentation method based on an improved convolutional neural network is proposed. First, the residual learning is combined with the densely connected network (DenseNet) to fully exploit the feature maps of each layer. The path from the low-level feature maps to the high-level ones via the addition of shortcuts is shortened and the feature propagation ability is strengthened. Second, as for the extraction of more fine vessels, the dilated convolutions are adopted in the encoder-decoder network to expand the receptive field without the increase of parameters. The experimental results show that the proposed network structure has less parameters, compared with the other existing deep learning methods. The average accuracy on the DRIVE datasets is up to 0.9556, the sensitivity is up to 0.8036, the specificity is up to 0.9778, the area under curve of receiver operating characteristic reaches 0.9800, better than the segmentation effects of the other existing deep learning methods.

The primary mirror of large aperture solar telescope is directly exposed to the air, the temperature difference between the primary mirror and the environment causes the mirror seeing. In order to decrease the mirror seeing, we study the influences of the temperature and the velocity of the cooling air as well as the velocity of the air knife on the mirror seeing. Combined with the primary mirror of 2.5 m called Imaging and Spectroscopic Multi-Application Telescope (ISMAT) of Nanjing University, thermal control technology is studied, and the optimization goal is mirror seeing of 0.02 arcsec. The results of numerical solution show that: the cooling air temperature and the inlet velocity satisfied the optimization goal are 283.15 K and 3.5 m/s without the air knife; the cooling air temperature, the inlet velocity, and velocity of air knife satisfied the optimization goal are 283.15 K, 3 m/s and greater than or equal to 3 m/s.

In the three-dimensional measurement of structured light, the measurement structure only with a single projector is prone to occlusion and shadow in the measurement of complex objects and thus the integrity of measurement data is influenced. In contrast, the measurement structure with double projectors can broaden the imaging area and reduce the shadowing region, and thus the measurement efficiency is improved. However, it is still difficult to separate the original projection signals from the partially overlapping phase-shift gratings for this method. On the basis of the analysis of the conventional four-step phase-measuring method, a separation method of overlapping phase-shift gratings is proposed via the complementarity of gray scales, in which the invalid overlapping signals are effectively eliminated by adjusting the time and sequence of the projected phase-shift gratings while keeping the original measuring structure. As for the verification of the effectiveness of the above separation method, a measurement system is established to conduct the related experiments. The results show that the complementarity of gray scales can be used to realize the accurate separation of overlapping phase-shift grating signals, obtain the complete point cloud data, and improve the measurement speed.

For rigid objects moving on-line on the pipeline, the projection composite grating can solve the contradiction between different frequency requirements for pixel matching and phase unwrapping. However, when the phase is calculated, it is necessary to filter the composite grating, which will reduce the reconstruction accuracy. Based on Stoilov algorithm, an on-line three-dimensional(3D) measurement method of composite grating projection without filtering is proposed. By the design of the composite fringe, the phase shift direction of the low-frequency fringe is parallel to the moving direction of the measured object, and the motion of the measured object after pixel matching is converted into the phase shift of the low-frequency fringe. The phase shift direction of the high-frequency fringe is perpendicular to the moving direction of the measured object, and the light intensity distribution of the high-frequency fringe in each frame deformed fringe pattern after pixel matching is exactly the same. Therefore, the phase calculation can be performed directly, and the loss of accuracy is avoided due to filtering. Meanwhile, the intensity of the high-frequency fringe in the composite grating is much lower than that of the low-frequency component, so the high-frequency fringe can be regarded as a weak background light. The accuracy of on-line 3D measurement is guaranteed. Simulations and experiments prove the proposed method’s validity.

In order to solve the problems of overexposure or underexposure that are easily encountered when reconstructing a strong reflective surface in visual measurement, we propose an automatic multiple exposure surface structure light measurement method. The core of this method is to calculate the camera response curve according to the variation of the reference point pixel value with the exposure time. Using camera response curves and images at different exposure time, we calculate the numbers of exposure and exposure time required for the current scene measurement. Image sequences with different exposure time are fused into a new fringe image sequence for reconstruction. The experimental results show that the method can accurately calculate the time of each exposure and overcome the problem of saturation or oversaturation of the fringe image caused by the strong reflection surface, and realize the three-dimensional optical non-contact measurement of the object with high dynamic range surface reflectivity.

A model of the image highlight removal based on the back propagation neural network and the Gaussian distribution functions is established, which is beneficial to optimize the image feature extraction and the image feature matching. With the high reflectivity workpiece as the experimental object, the extraction of edge features and the analysis of the vision measurement accuracy are conducted. The experimental results show that the proposed method can be used to realize a vision measurement accuracy of 0.75 mm, which verifies the feasibility of this proposed method to a certain degree.

Geometric calibration errors in deflectometry are the main factors limiting the precision of low-order surface measurement. The relationship between the geometric calibration errors and low-order surface measurement errors is analyzed with mathematic model, theory simulation and experiment. The mathematic model for the relationship between the geometric calibration errors and the surface measurement errors is introduced and verified by simulations and experiments. The results show that the coordinate translational errors in geometric calibration introduce tilt and defocus into the surface measurement results. Besides, the longer distance between the tested mirror and the camera, display, the less effect the geometric calibration errors have on the low-order surface measurement. The results can help designing the system configuration of deflectometry and improving the accuracy of low-order surface measurement.

The laser output spectral characteristics are measured and studied with the discharge initiated pulsed HF laser developed by our project team. The results show that the spectral energy of the discharge initiated pulsed HF laser is mainly concentrated in three vibrational transition bands of v(3—2), v(2—1) and v(1—0). The 13-16 main output spectral lines are obtained by the measurements in the wavelength range of 2.65-3.05 μm, in which the v(2—1) vibrational transition band has the highest output energy ratio, about 40%-50%. The output energy ratio in each vibrational transition band can be tuned via the suitable adjustment of charging voltage and total working gas pressure. The output energy ratio in the v3 vibrational transition band increases with the increase of laser charging voltage, but decreases with the increase of total working gas pressure, which is opposite to those for the v1 vibrational transition band.

By taking the single checkerboard image of fisheye cameras to characterize as targets and comprehensively using multiple geometric constraints, the initial values of fisheye camera parameters are solved in different steps and the global optimization is simultaneously implemented. The exact principal point location (u0, v0) of the camera is obtained by means of fisheye image contours and their symmetry, in which the difficulty in detecting contour points under a black background is skillfully avoided through scanning the bounding box of fisheye image contours. Two group of projection ellipses on the fisheye checkerboard images are accurately fitted whose intersection points are back-projected to the unit sphere to obtain vanishing points of the parallel lines, and the initial value (fx, fy) of the equivalent focal length of camera and the initial angle of rotation matrix are deduced from the orthogonal constraints of two vanishing points. With the radial alignment constraint and the checkerboard corner point information, the initial values of the translation vectors (tx, ty) are first solved linearly, then that of tz is obtained after the establishment of a quadratic equation with one unknown, and finally, through minimizing the re-projection error of the checkerboard corner point, all the camera parameters except for the principal point are globally optimized and these optimized parameters are used for the correction of fisheye images. Two focus-fixed Hikvision fisheye cameras with different fields of view are selected for the calibration and image correction tests. The results show that the re-projection root mean square error (RMSE) of each camera for the proposed method is less than 1/3 pixel, and the calibration parameters can keep a good stability in the planar perspective correction effect at different areas of the fisheye image with the correction effect in the central region slightly better than that at the edge. The RMSE in the checkerboard corner point line fitting of the corrected fisheye images is less than 0.7 pixel which is obviously superior to that from the online calibration toolbox, and thus it has a relatively good application value.

Aiming at the problems of object tracking failure caused by occlusion and out of view in long-term tracking, we propose a long-term object tracking algorithm based on feature fusion to improve the speed and robustness of object tracking. First, the features of histogram of oriented gradient, color space and local sensitive histogram are fused to enhance the robustness of the algorithm in complex cases, and the fusion feature dimension reduction is carried out to improve the object tracking speed. Then, an additional one-dimensional scale correlation filter is used to obtain the optimal scale estimation of the object, and the computational complexity is reduced by quadrature rectangle-factorization. Finally, the object detection threshold is adaptively determined. When the object occlusion or out-of-view causes the failure of object tracking, the object region proposals can be extracted by EdgeBoxes, and object position is re-directed by using structured support vector machine to complete the long-term tracking of object. Experiments are conducted on standard tracking datasets OTB2015 and UAV123. Experimental results show that the average accuracy of the proposed algorithm is 5.0% higher than that of other optimal algorithms, the average success rate is increased by 2.6%, and the average object tracking speed is 28.2 frame/s, which meets the real-time requirements for tracking. In the case of object occlusion and out of view, the proposed algorithm can track the object continuously and accurately.

Recently, the framework of convolution neural network has been successfully applied to the target tracking, and has achieved robust tracking results. On the basis of this conception, a target tracking method based on location-classification-matching model is proposed. First of all, in the location model,the candidate target region of the current frame is predicted by using location information of previous frame. Secondly, the trained deepth features are used to inter-class screen the candidate regions, and N sub-optimal target regions are selected. Finally, we use conventional color features to perform intra-class optimization matching for sub-optimal target regions, so as to determine the final tracking target. Meanwhile, the network in the location and the classification is updated separately, and the established target model is updated online and real-time to ensure that the model describes the target accurately. Experimental tests are performed on OTB50 and OTB100 standard databases, the experimental results show that the proposed tracking method has better tracking robustness under the conditions of fast motion, similar object interference, and complex background.

The picture quality is declined in the low illumination environment. Meanwhile, the fog and haze formed by smoke, dust and other substances suspended in the air will cause blurred image details, which have a great impact on outdoor photography and computer vision. Therefore, it has important application value for image processing and computer vision by defogging degraded images to improve image quality. We propose an image defogging algorithm based on the combination of bright and dark channel in fog and haze weather. A model of the air light scattering is proposed based on the physical model of degraded image. Air light value and transmissivity are estimated by using the combination of light channel prior and dark channel prior. The algorithm can solve color distortion problem of sky area when fog-free image is restored, recover the image details and color, and improve the vision effect of the image. Evaluation parameters are used to compare the image quality. Simulation results show that the algorithm proposed in this paper is better than multi-scale Retinex image defogging algorithm.

In order to estimate the calibration error, analyze the factors affecting the calibration error, and guide the camera calibration system engineering design before the calibration system design, we propose a new algorithm model for the P4P camera calibration algorithm. This model solves the problem that the azimuth, pitch angle and roll angle are coupled to each other, and makes the attitude angle solution only related to the camera internal parameters and the image coordinates of feature points. On this basis, the model of error analysis is built which can theoretically analyze the influence of image position errors, aberration, principal point, focal length, and locations of feature points on calibration of attitude angles. The simulation and experiment are carried out. The conclusion hints that the data of simulation and experiment of calibration accuracy are consistent and the error analysis model is accurate and valid. The model can guide the design of calibration system and has great engineering application value.

We propose an algorithm that solves the perspective-n-point problem in the framework of variable projection. The perspective-n-point is the problem of determining the rotation and translation parameters of a camera, which is widely used in computer vision, photogrammetry, robotics, space rendezvous and virtual reality. We separately estimate the rotation and translation parameters in the variable projection framework. Firstly, an optimal rotation matrix is solved by using the Wahba problem with fixed translation parameters. By applying the formula for the derivative of the singular value decomposition, the derivative of the optimal rotation matrix is calculated with respect to the translation parameters, then the derivative of the objective function with respect to the translation parameters is obtained. Finally, the objective function is minimized by Levenberg-Marquardt algorithm over the translation parameters. The experimental results show that the proposed algorithm is one of the most accurate and effective algorithms and has a large convergence basin.

To solve the problem of low matching accuracy in textureless regions, a local stereo matching method is proposed based on improved cost computation and adaptive shape guided filter. First, an efficient cost function combining enhanced image gradient and enhanced gradient-based Census transform is introduced for cost computation. Then, an adaptive shape cross-based window is constructed for each pixel, and guided filter aggregation is implemented based on this adaptive window. The final disparity map is obtained after disparity computation and multi-step disparity refinement. The experimental results demonstrate that the average matching error rate of the proposed algorithm is 4.80% for stardard image paris on Middlebury benchmark. Compared with traditional guided filter-based method, the proposed method has better matching results in textureless regions.

The formation reasons and suppression methods of the typical defects generated on the surface of the potassium dihydrogen phosphate (KDP) crystal are investigated. The correctness of the analysis results of the formation reasons are confirmed via the fly-cutting fabrication and the surface painting fly-cutting experiments, and the formation process of the surface defects of the KDP crystal are further clarified. Theoretical models which are suitable for the description of the formation process of the surface defects of the KDP crystal are established and the process conditions for the realization of the defect-free KDP surfaces are shown. The fly-cutting fabrication parameters and the cutting tool structures are optimized, and thus the validity of the suppression methods is verified. The research results show that, the variance range of the brittle ductile transition (BDT) depth of the (001) plane of KDP crystal is 125-268 nm under the fly-cutting conditions. When the cutting is along 45° direction, the BDT depth reaches the maximum. At this point, the fracture pits can be avoided on the crystal surface as long as the feed rate does not exceed 36.6 μm·r-1. Besides, the convex defects are eliminated by the optimization of the cutting tool structures, the surface defects of the KDP crystal can also be suppressed effectively, and thus a smooth KDP surface with a roughness less than 2 nm is obtained.

The narcotics such as ecstasy, methamphetamine, and heroin normally can emit fluorescence under excitation with certain wavelengths. Because of the small number of fluorescent groups in the molecular structure or the low quantum efficiency, the fluorescence intensity of narcotics is relatively weak and the narcotics with low concentration are difficult to be detected, which severely hampers the detection of the narcotics. In this paper, for improving the detection sensitivity of the fluorescence spectrum technology, the divergent characteristics of fluorescence signal are considered to increase the collection efficiency, which is different from traditional methods such as choosing high power laser to increase the excitation light energy and selecting high sensitive detector. In the optical path, parabolic mirror and Fresnel lens are used to collimate the fluorescence light, while the plano convex lens and double convex lens are used to contract the fluorescence beam, and the beam is focused by a certain size micro-lens array. Simulation results show that the fluorescence collection efficiency of the optical path is enhanced up to about 21.08%, which is six times more than that of the traditional collection method. The research work provides a technical reference for improving our portable and high sensitive narcotics fluorescence detection system.

Photoacoustic imaging has great potential in the field of biomedical imaging because it is a beneficial combination of the high contrast of pure optical imaging and the high resolution of ultrasound imaging for deep-tissue. By acquiring the photoacoustic signals at multiple locations, we can obtain a two-dimensional or three-dimensional optical absorption distribution image of the biological tissue. However, it is difficult for actual photoacoustic imaging to acquire the photoacoustic signals with enough detector locations due to the constraints of hardware conditions and imaging time. In the case of insufficient signal sampling, the reconstruction quality of the photoacoustic image is seriously degraded, and a large number of artifacts appear consequently. To overcome this problem, we propose a reconstruction strategy which uses photoacoustic signals preprocessed by a recovered algorithm based on dictionary learning and sparse representation, and simulation experiments are carried out. The results show that by applying the proposed algorithm, a photoacoustic image can be reconstructed with less artifacts, clearer details and 8 dB peak signal-to-noise ratio improvement compared with images reconstructed without super-resolution reconstruction of photoacoustic signals. The simulation experiments with different signal-to-noise ratios verify that the proposed algorithm has good robustness.

A surface-emitted ring-cavity terahertz (THz) wave parametric oscillator is constructed based on the 1% molar fraction of MgO-doped near-stoichiometric lithium niobite crystal, whose output THz frequency tuning range is 0.99-3.84 THz and frequency tuning response time is 600 μs. When the pump pulse energy is 150.30 mJ and the THz frequency is 1.59 THz, the output energy of the THz pulse is the maximum 16.28 μJ and the corresponding energy-conversion efficiency is 1.08×10 -4. Moreover, under the same experimental conditions, the maximum THz wave output energy from this ring-cavity THz wave parametric oscillator is about 2.35 times that obtained from the conventional linear-cavity ones. The THz wave output with high-energy and rapid-tuning is realized.

The digital controller model of the primary mirror control system in an 8 m segmented ring solar telescope (8-m-RST) is established. The relationships between the sampling period, relative stability, integral gain and the control bandwidth are obtained by means of the extraction of the frequency characteristic parameters from the system model. In addition, with the introduction of a pulsed wind disturbance model, the performance of the primary mirror system under the disturbance of a pulsed wind with a low average speed is tested by simulation. The research results show that, the primary mirror control system in the 8-m-RST is stable and its control width meets the requirement of 0.2 Hz. Moreover, it can effectively suppress the disturbance from a wind with an average speed of 2 m·s-1. These results have important references to the structural improvement and the design of the tilt sensor and the controller in the 8-m-RST.

A selection method for the sub-band stitching optical glass combination is proposed. The apochromatic equation is solved by the least square method, and the optimal glass combination which can effectively correct the tertiary spectrum is obtained. A broadband long-focus optical system is designed based on this optical glass combination with a wavelength range of 0.45-1.014 μm, a focal length of 400 mm, an F number of 6, and a full field of view of 10°. This system consists of three kinds of optical glasses and seven spherical lenses, and the modulation transfer function (MTF) of this system is closed to the diffraction limit. The research results show that the calibration of the tertiary-spectral residuals in the apochromatic optical system can be achieved with this proposed method.

A spatial optical detection system with a large field of view is designed. The parameters of this system are calculated by combining the spatial target characteristics and the performance indexes of the detector. Thus, the technical indexes of this system are determined, and the detection of the ninth-magnitude star target 255 km away in space is realized. This system adopts an off-axis three-mirror optical structure with a spectral range of 400-900 nm, a focal length of 64 mm, and a field of view of 30°×30°. The primary and tertiary mirrors in this system are designed based on the Zernike and XY polynomial freeform surfaces. The ray tracing is performed and the discrete point data are obtained. These data are fitted by MATLAB to obtain the profiles of the freeform surfaces of these mirrors. This system has a high energy concentration and a good imaging quality.

An optical system for the non-mydriatic stereoscopic imaging fundus camera with a field of view of 30° is designed, which is composed of an imaging system and an illumination system. In the imaging system, a new optical structure for the stereoscopic imaging of fundus is designed, in which a prefixing objective lens is introduced to improve the imaging resolution. In the illumination system, an annular aperture is introduced to avoid the generation of the corneal reflected light and the black spot plates are adopted to eliminate the stray light produced by the ophthalmic lens. The research results show that, this system can be used to achieve not only the multi-angle synchronous acquisition of the retina images, but also the highly clear imaging of the fundus with 6×106 pixel. As for this system, the object resolution of the normal human eyes is higher than 200 lp·mm-1. This system has an overall length of 290 mm, a field curve value of smaller than 28 μm, and a distortion value of only -4.9%. This system has a strong focusing ability and the human fundus with a diopter range of -7-+5 m -1 can be imaged clearly.

Ground-based photoelectric detection system is an important detection method for signal acquisition of space targets, and exposure time is the main parameter affecting its detection performance. Based on the optical properties of the space target, limit detection distance, minimum detectable size and limit detection star of the system are calculated, and the quantitative relationship between the exposure time and other influencing factors is analyzed. Combined with the image shift model of the space target, the relationship between the system detection ability and the exposure time under dynamic conditions is obtained, and a simulation study is conducted. The results show that the system detection capability increases with the exposure time and becomes stable when the target is relatively stationary. When the target is in relative motion, the system detection capability increases firstly and then decreases with the extension of the exposure time, and there is an optimal exposure time related to the target relative angular velocity. The simulation results can provide optimization feasibility for the exposure time setting of ground-based photoelectric detection system.

Rapid and accurate assessment of the primary productivity of phytoplankton is crucial to scientific research of marine ecological environment and understanding of evolution rules of global carbon cycle. The measurement period of primary productive forces represented by the 14C tracer method or the black and white bottle method is long and cumbersome. To solve the problem, we study the photosynthetic electron transfer rate according to the biofilm energy flow theory. Photosynthetic parameters are obtained by the variable chlorophyll fluorescence induced by light source, and combined with photosynthetic electron transfer rate “biological-optics”. Therefore, the measurement of photosynthetic electron transfer rate based on fluorescence kinetic parameters is studied. The chlorella pyrenoidosa is tested under different stress conditions, and the photosynthetic oxygen rate measured by the liquid phase oxygen electrode is compared to verify effectiveness of the photosynthetic electron transfer rate. The results show that under different concentrations of dichlorophenyl dimethylurea stress, the photosynthetic electron transport rate and photosynthetic oxygen release rate of phytoplankton have good consistency, and both decrease significantly with the increase of the stress concentration. The photosynthetic oxygen release rate and photosynthetic electron transport rate are reduced by 71.55% and 68.87%, respectively, and their square of correlation coefficient is 0.934. Under different nutrients or light intensity for 15 d, the photosynthetic electron transport rate and photosynthetic oxygen release rate still have good consistency, and their square of correlation coefficient is greater than 0.955.

The principle of generating the spiral Bessel beam is theoretically analyzed. Based on the cross-spectral density function of Gaussian-Shell model and the theory of cross-spectral density propagation, the expression of partially coherent spiral Bessel beam is derived. The propagation characteristics of partially coherent spiral Bessel beam are studied and the influence of coherence on its transmission characteristics is analyzed. The results show that with the increase of transmission distance, the contrast of beam intensity distribution of beam cross section decreases, and the distribution of light field changes from the Bessel type to the Gaussian type. However, the beam still maintains the self-acceleration property. For the high-order beam, a phenomenon of dark core fading can be found easily and the fading distance of the dark core increases with the increasing of the beam order. With the decreasing of the coherent length, the distribution of Bessel light field decreases gradually, and the non-diffraction distance of the beam is shortened. Moreover, the influence of light field regulation on the beam is also discussed, which includes the parameter setting of the hologram and the adjustment of the coherence of the light field.

A scene classification algorithm of remote sensing images based on the integrated convolutional neural network (CNN) is proposed. A back-propagation network is constructed to measure the complexity of scene images. The classification of these images is conducted with the CNN based on the complexity level of each image, thus, the scene classification of remoting sensing images is achieved. With the proposed algorithm, the experimental verification of the open data of NWPU-RESISC45 is conducted and the classification accuracy of 89.33% for Type I test and that of 92.53% for Type II are obtained, respectively. The average running time is 0.41 s. Compared with the VGG-16 model for fine tuning and training, the classification accuracy by the proposed algorithm is increased by 2.19% and 2.17%, respectively. Simultaneously, the prediction rate is increased by 33%. Thus, the efficiency and practicality of this proposed algorithm are confirmed.

A fast algorithm based on the boundary of laser beam is proposed. Based on the time-space distribution of laser radar signals, the response function of the detection target to the signal and the model of the interaction process between them are deduced and established. A target reconstruction simulation system based on the laser reflection tomography is constructed. Based on the influences of the different sampling intervals and the different detection angles on the reconstructed image quality, two sets of comparative simulation experiments are carried out. Under different conditions, the filter backprojection algorithm is used to achieve the image reconstruction of the two-dimensional contour for the target.

Ammonia (NH3) is one of the main precursors of secondary fine particulate matters in the atmosphere. The accurate measurement of NH3 concentration plays an important role in the monitoring and protection of atmospheric environment. Although the cost of a near-infrared laser is relatively low, there always exist some problems such as the H2O, CO2 gas interference in the environment and the limited absorption path when it is used for the NH3 concentration measurement. In order to reduce the H2O and CO2 interference in the environment, we screen out the absorption line with a central wavenumber of 6521.97 cm-1, which is used for the measurement of the trace NH3 concentration in the atmospheric environment. Moreover, this spectral line is not affected by the CO2 absorption in the environment and its overlap range with the H2O absorption line is small under the condition of low pressures. The spectral absorptivity of NH3 can be accurately extracted through multi-peak fitting. A measurement device based on cavity ring-down spectroscopy is built with a distributed feedback laser, in which the ring-down cavity is composed of a pair of high reflectivity mirrors with a reflectivity of up to 99.996%, the empty ring-down time is about 96 μs,and the effective absorption path is about 1.6×104 m. The trace NH3 concentration in the atmospheric environment is measured with this device, and the results show that the detection sensitivity of this measurement system can approach 3.9×10-10.

In order to quantify the spectral data of a miniature infrared spectrometer based on hollow planar waveguide, we built a wavelength calibration device based on carbon dioxide (CO2) laser and integrating sphere on the basis of analyzing the common wavelength calibration method used in the laboratory. The spectrum calibration range of the device is wide and the resolution is high, which overcomes the shortage of traditional laboratory calibration methods and improves the calibration precision. We first introduce the working principle of the infrared spectrometer based on hollow planar waveguide, then use the calibration device to calibrate the wavelength of the spectrometer, and then complete the calibration data analysis by the polynomial fitting algorithm. Finally, the calibration results are verified by the measurement of two narrow band filters. The experimental results show that the calibration error of the central wavelength is not more than 0.02 μm and the spectral resolution can reach 144 nm by using the combination of CO2 laser and integrating sphere.

When spatial modulation Fourier transform spectrometer is used to detect the spectrum of remote target, the interferogram and recovered spectrum are influenced by wavefront distortion resulting from atmospheric turbulence disturbance. According to the phase modulation of atmospheric turbulence disturbance on optical field, we build the model of atmospheric turbulence random phase screen and optical field split-step propagation in atmosphere. The interferogram and recovered spectrum affected by atmospheric turbulence disturbance are calculated numerically. The results show that the atmospheric turbulence disturbance causes low-frequency background intensity fluctuation in interferogram, and the concomitant frequency noise appears at the low-frequency region of the recovered spectrum. The relationship between normalized spectrum error and telescope entrance pupil magnification along with atmospheric coherence length is analyzed by statistical experiment method. The results indicate that the statistical mean of normalized spectrum error is linear positive correlated to telescope entrance pupil magnification, and it is nonlinear negative correlated to atmospheric coherence length. According to the statistical result of the normalized spectrum error, in order to realize the effective detection on target spectrum, the telescope entrance pupil magnification can be designed reasonably on the basis of the atmospheric coherence length in outfield environment.

The tunable diode laser absorption spectroscopic technique is used for the measurement of water vapor in high purity nitrogen. Firstly, in the self-designed multiple reflectors, the accurate light path values of the multiple reflectors are measured by using the method comprising of taking CH4 with a volume fraction of 1.007×10-3 as the standard gas, the methane absorption spectrum at around 1654 nm, and the three-line simultaneous fitting of three absorption lines with the similar central frequencies (less than 0.01 cm-1) and the same low energy levels. Secondly, the laser interference background and the absorption background of ambient air in the tunable diode laser absorption spectroscopy system are studied and the accurate background of tunable diode laser absorption spectroscopy system is obtained. Finally, the water vapor concentrations in high purity nitrogen are measured by the 1854 nm tunable diode laser absorption spectroscopy system with a detection sensitivity of 1.14×10-6. Through the strict background absorption deduction, the multi-line Voigt linear model is used to fit the water vapor spectrum without background absorption to obtain the water vapor concentration in high purity nitrogen. The results show that, the obtained maximum deviation is 10.33% compared with the national standard.

Based on the combination of the three-dimensional fluorescence detection technique and the alternating penalty trilinear decomposition algorithm, a method is proposed to detect oil species in mixed oils. First, three-dimensional fluorescence spectra of 20 samples of the mixed oil (jet fuel and lube) with different volume ratios are obtained by the FLS920 fluorescence spectrometer, and calibrated by the Delaunay interpolation method. Then, the number of components required for the alternating penalty trilinear decomposition algorithm to analyze three-dimensional fluorescence spectra is determined by the core consistency function. Finally, the effectiveness of the alternating penalty trilinear decomposition algorithm in analyzing three-dimensional fluorescence spectra is evaluated by the root mean square error and the correlation coefficient matrix. The results show that both of the root mean square error and the non-diagonal elements in the correlation coefficient matrix meet the threshold requirements of 0.05 and 0.95, which are obtained by the alternating penalty trilinear decomposition algorithm. As for the solution of the serious overlapping problem of three-dimensional fluorescence spectra, the alternating penalty trilinear decomposition algorithm is superior to the parallel factor algorithm. The purpose of rapid detection of oil species in mixed oils can be achieved by the alternating penalty trilinear decomposition algorithm.

The metal coating material of Ag and dielectric thin film materials such as Ta2O5, SiO2 and Al2O3 are selected for designing and fabricating a kind of polarization-maintaining mirror on the quartz substrate, whose polarization contrasts at wavelengths of 810 nm and 850 nm are superior to 10000∶1. With the combination of the orbital environment for the polarization-maintaining mirror, the experiment for simulating the spatial atomic oxygen is conducted and the change law of polarization contrast of the polarization-maintaining mirror sample is investigated. The results show that, with the increase of the dose of the atomic oxygen, the polarization contrast of the polarization-maintaining mirror shows an overall decreasing trend. Moreover, the decreasing trends along the directions of +45° and -45° are more obvious than those along the horizontal (H) and vertical (V) directions.

Based on the description of Compton scattering cross section by Klein-Nishina(K-N) formula and combining with Beer law, the probability of scattering distribution caused by each scattering point set in the object to all detector elements is calculated respectively, and the probability of scattering distribution caused by superimposing all scattering points is calculated. The total scattering distribution is obtained by adjustment of the coefficients. Finally, the scattering distribution is reduced from projection data, and the correction of scattering artifacts is achieved. The results from simulation and experimental show that the proposed method can significantly suppress the artifacts and shadows with shape of cup caused by scattering and improve the quality of reconstructed images.