By using the sun photometer, nephelometer and particle matter monitor of national climatology observatory in Shouxian, the atmospheric optical characteristics in Shouxian are observed and analyzed in winter. The daily variation rules of optical depth of aerosol, scattering and absorption coefficient, and single scattering albedo are obtained. Results show that, with the increasing of aerosol content, particle matter content also increases, the optical depth of aerosol, scattering coefficient and single scattering albedo increase, and the aerosol's wavelength index and atmospheric turbidity coefficient also increase. The size changing of aerosol particles can be characterized by asymmetric factors, backscattering ratio, scattering coefficient, single scattering albedo and wavelength index together.

By exploring the range errors from the target specular reflection, and analyzing the relationship between range errors and intensity data, a range error model from the target specular reflection based on original intensity data is established. Experimental results show that the original intensity data can be used to establish a range error correction model, which is independent of target material and surface geometry. According to the original intensity, the distance data can be accurately corrected, which is of great significance to improve the accuracy and quality of terrestrial laser scanning.

The cluster-cluster aggregation model of soot and the discrete dipole approximation method are used to analyze the single and multiple scattering channel characteristics of ultraviolet (UV) light under different soot concentrations and particles sizes as well as the influence of scattering angle on the scattering intensity. The research results show that with the UV line-of-sight communication method, the path loss increases with the increase of the soot concentration when the particle radius is the same, meanwhile when the soot concentration is constant, the path loss also increases with the increase of particle size. With the UV non-line-of-sight communication method, the larger the soot concentration and the original radius, the less the path loss of the scattering channels at a short distance. In contrast, at a long distance, the higher the soot concentration,the larger the path loss, however, the difference among path losses for different concentrations is small. In the case of multiple scattering, the scattering light intensity decreases with the increase of scattering angle, while when the scattering angles are the same, the higher the soot concentration, the greater the scattering UV light intensity.

Based on the standard nonlinear Schr dinger equation, the first-order breather solution and its rouge wave limit are obtained with Darboux transform method, and the dynamic characteristics of first-order breather solution are studied. High-order breather solutions of nonlinear Schr dinger equation are obtained by means of recurrence relation of Darboux transformation. And their collision superposition, separation, degeneracy and parallel transmission modes are studied, respectively. Nonlinear Schr dinger equation's rouge wave limit can be obtained when each breather frequency tends to zero. Research results show that the rouge wave's amplitude, number of bumps, order of center peaks and number of surrounding peaks after splitting are related to rouge wave's order.

The light guiding mechanism of negative curvature hollow core fiber is analyzed based on anti-resonant reflected optical waveguide model and mode coupling model. Influence of structural parameters of circular tube negative curvature hollow core fiber on confinement loss of fundamental mode is simulated based on finite element method. Reduction reason of confinement loss of fundamental mode with nested tubes is considered, which is found that air layer between nested tubes has contribution to anti-resonant reflection, and nested tube has different potentials for confinement loss optimization under negative curvature fiber structure with different number of tubes. A negative curvature hollow core fiber with core diameter less than 20 μm is proposed with confinement loss less than 1 dB·km -1 at 1550 nm, and single mode transmission can be maintained.

A carbon monoxide (CO) gas sensor based on copper oxide/polyaniline composite membrane-coated photonic crystal fiber is proposed. The Mach-Zehnder interference structure is formed by fusion splicing a standard single-mode fiber to a solid photonic crystal fiber. The copper oxide/polyaniline composite material is coated on the surface of the photonic crystal fiber, and then the purpose of detecting carbon monoxide can be achieved. The results show that a uniform layered film is formed on the surface of the fiber with a thickness of about 2 μm. A high sensitivity of 17 pm and a good linear relationship are achieved in the range of CO volume fraction of 0~75×10 -6. The response time and recovery time are 80 s and 110 s, respectively. The sensor has an advantage of low cost, simple structure, and easy to manufacture.

In order to realize the quasi-real-time calculation of holograms, a fast algorithm of computer generated holograms based on a multi-graphic processing unit (GPU) in Matlab is proposed. This method makes full use of the simplicity of Matlab programming and the high computational performance of GPU, and thus the computing time of holograms is effectively saved. The simulation results show that the computing speed of the proposed method is about two orders of magnitude higher than those of the traditional calculation methods.

Aiming at the existing problems in the line detection for standard Hough transform, a line detection algorithm based on improved random Hough transform is proposed. The pixels of edge images are clustered and grouped by 8-neighborhood search. The concept of the pixel gradient direction difference is proposed, and the gradient direction difference between adjacent pixel is calculated in each edge group, thus the line pre-detection is carried out to exclude the edge groups without line features. Based on the theory of the random sample consensus algorithm, the improved random Hough transformation algorithm with a linear parameter pre-test model is proposed. The research results show that the proposed algorithm effectively solves the problem in standard Hough transform and improves the error detection rate in the process of line detection. The proposed algorithm has the advantages of fast detection and high detection accuracy.

Aiming at the problem of object detection in remote sensing images, the Faster-Rcnn network based on the convolutional neural network models is used to extract the features of the object area. An object detection dataset containing three kinds of common targets in remote sensing images is made to train this network. In addition, in order to solve the problem of large rotation angle of remote sensing images, a target detection model with a rotation invariance self-learning ability is proposed, which integrates the spatial transformation network into the Faster R-CNN framework. By the analysis and comparison with the traditional object detection methods, the true effects of object detection in remote sensing images by different methods are explored. The features extracted by the convolutional neural networks based on the spatial transformation networks possess stronger orientation robustness than those by the traditional methods, which makes it possible to obtain a high detection precision.

In order to evaluate the camouflage effect of national defense engineering using hyperspectral remote sensing images, a comprehensive evaluation method for hyperspectral camouflage effect based on intuitionistic fuzzy decision is proposed. A comprehensive evaluation index system for hyperspectral camouflage effect is established, which including spectral pan-similarity measure, brightness contrast, visual similarity measure and structural similarity measure. A multi-attribute decision-making model based on intuitionist fuzzy sets is established for hyperspectral camouflage effect evaluation as well. The algorithm for order preference by similarity to an ideal solution is adopted for model decision, and a grey correlation measure based on Hamming distance is further proposed to improve the distinguishing ability between different alternatives. The research results show that the evaluation results obtained by the proposed method are consistent with those from expert evaluation, and thus its feasibility and rationality are verified.

The traditional image recognition algorithm has only a single recognition model and is susceptible to the external illumination interference. In contrast, as for the deep convolutional network model, there exist a large amount of calculation and high cost although its recognition rate is high. An improved based compression algorithm is proposed based on the deep XNOR-network. The compositions of the weld recognition system and the classical convolution neural network model are first introduced. The improved convolution network compression algorithm is described, including the weight update algorithm and the weight compensation algorithm. The data experiments are performed on the self-made datasets and the simulation platform. The research results show that the proposed algorithm has the advantages of high recognition rate, small model, strong adaptability and diversity of recognition models, which can be applied to the weld identification in the welding site.

Aiming at the problem that the current approaches of semantic segmentation cannot meet the simultaneous demands on accuracy and efficiency in scene parsing in the intelligent vehicles, an accurate and efficient algorithm for semantic segmentation is proposed. Based on the proposed separable residual module and the down-sampling module, a real-time and accurate semantic segmentation network is designed. With the Cityscapes dataset, the segmentation accuracy can reach 67.86% on the basis of the 12 frame/s efficiency. The research results demonstrate that the proposed method can achieve a good performance both in accuracy and efficiency, and makes a balance between accuracy and efficiency.

Based on the efficient FAST (features from accelerated segment test) algorithm, a color based-FAST (C-FAST) algorithm for image feature detection and matching with color information improvement is proposed. The algorithm adds scale invariant and rotation invariant feature descriptors which the original FAST algorithm does not have, and takes color information as an important reference factor in feature detection and matching. Hence, the proposed algorithm is more efficient and has higher detection and matching accuracies. It also has good robustness under the conditions of illumination changes and noise effects. Different algorithms like FAST, speeded up robust features (SURF), colorful scale-invariant feature transform (CIFT) and the proposed algorithm are analyzed via public data sets and common images. The running data prove that the proposed algorithm can detect and match the image features effectively and reliably, with 30% accuracy improvement compared with the original FAST algorithm.

In order to explore the influence of human disturbance activities on soil organic matter content and improve the estimation accuracy of soil organic matter in arid area. The soil in Fukang City, northern Xinjiang, is studied. The hyperspectral curves of 90 sampling points are successively transformed by continuous wavelet transform (CWT) and compared with R′and lg(1/R), two common spectral transformation methods. The results show that the spatial variability of soil organic matter is enhanced with the increase of human disturbance degree. Moreover, in the common spectral transformation methods, R2 of partial least squares models established by the soil organic matter institute and R′ in zone I, zone II and zone III are both higher than those by R and lg(1/R). The precision of the model established after CWT is very high. Compared with R2 of the model established by R, R2 of the measured value and the predicted value are 0.717, 0.689, 0.630, increased by 0.382, 0.4, 0.389, respectively. In addition, the relative percent deviation correspondingly reach 2.150, 2.090, 2.013, indicating that CWT can well predict the soil organic matter contents. The usage of CWT does not greatly reduce the model accuracy with the increase of human disturbance degree and is more suitable for the prediction of organic matter content in arid regions.

A three-channel oceanographic shipborne lidar system is built based on a 532 nm pulse laser. The concentration of chlorophyll-a is measured with this system and the results are compared with that obtained by the commercial chlorophyll fluorometer. The comparison results show that, the correlation coefficient between these results as high as 0.84, which shows a good linear correlation. In order to test and assess the detection performance of this system on the chlorophyll-a concentration of the surface seawater, the preliminary field navigation testing is carried out and the distribution of the chlorophyll-a concentration of the surface seawater along the navigation path is obtained. The results show that, the correlation coefficient between the chlorophyll-a concentration of the surface seawater extracted from the lidar detecting data and the chlorophyll-a concentration of the surface seawater obtained synchronously by the commercial chlorophyll fluorometer can reach to 0.69.

In order to improve the accuracy of three-dimensional reconstructed high-reflective surfaces, a phase fusion algorithm is proposed based on binary phase-shifting encoding. The phase error caused by the intensity saturation is analyzed and detected based on the periodicity and symmetry of the binary phase-shifting projection patterns. The research results demonstrate that, compared with the saturation error detection method based on the illumination intensity estimation, the proposed method can more accurately detect the areas with saturation phase errors. Compared with the result before correction, the root mean square error of the corrected phase shows a 95.7% reduction in the saturation phase errors and a 96.1% reduction in the maximum phase error. Compared with the traditional detection methods, the root mean square value of the replaced saturation phase error is reduced by 16.7% when fewer measurement groups are used. The results show that the proposed method can effectively reconstruct the surface profile of high reflective objects. The proposed method doesn’t need nonlinear correction, and is of strong anti-noise ability.

On the basis of analyzing the current methods for large spacing axes detection at home and abroad, a method of axial consistency detection based on non-cooperative target image processing technology is proposed, which selects scenes with typical features in far field as non-cooperative targets arbitrarily, utilizes different optical sensors to acquire non-cooperative target images, and obtains axial consistency detection results by comparing the spatial position differences of non-cooperative targets in images. Experimental results show that this method can meet requirement of large spacing axis consistency detection. While the CCD imaging system is reliably clamped on the mechanical axis of the target to be detected, the consistency detection between the mechanical axis and the optical sensor axis can also be achieved. This method is not limited by the optical sensor's working band and avoids the disadvantages of large volume, heavy carrying and high requirements on environment for other detection methods, so it has broad application prospects.

Controllable polarization beam splitter is a new type of optical waveguide functional device, which plays an important role in improving the performance of integrated photon system or developing new applications. A novel controllable polarization beam splitter is proposed by utilizing the large birefringence effect and high tunability of liquid crystal, in which the multimode interference optical waveguide structure is adopted. The optical performance of the polarization beam splitter is simulated and analyzed by the beam propagation method. The results show that, the polarization beam splitter is easy to achieve a high extinction ratio and a low optical loss. The polarization extinction ratio is over 28.7 dB and optical loss is less than 0.024 dB for both transverse electric (TE) wave and transverse magnetic (TM) wave polarizations. And the output ports of TE wave and TM wave can be dynamically controlled by adjusting the orientation of the crystal axis of liquid crystal. This device has several advantages such as simple structure, ease for design and manufacture, ease for integration with other photonic device, which has a wide potential application prospect in the photonic integrated circuits.

The QAl-7 aluminum bronze powder is cladded on the Q235 carbon steel surface by laser cladding technique in order to improve the surface performances of marine steels. Microstructure and copper contamination phenomenon near the fusion line are also studied. The research results show that the aluminum bronze coating undergoes liquid-phase separation and decomposition of oversaturated solid solution. The microstructure mainly consists of copper-rich phase (Cu3Al, Cu9Al4) and iron-rich phase (AlFe3, AlFe). The liquid copper alloy in the molten pool wets and extends along the austenite grain boundary of the substrates. There occurs the copper contamination phenomenon in the heat-affected zone. The contamination depth of copper alloy is related to the width of fusion zone. The fusion zone is wider, the copper contamination is deeper. Thus, the depth of copper contamination can be controlled just via the reduction of the width of fusion zone by controlling laser parameters. The copper contamination cracks are further suppressed.

In order to study the dilution effect in the process of laser cladding, the temperature fields in the process of single laser cladding of 304 stainless steel under different scanning speeds are simulated. Based on the melting point of materials, the values of dilution ratio (DR) are calculated. The laser cladding experiment is conducted on the surface of 27SiMn steel. The height and width of the cladding layer and the depth of the heat affected zone are measured, and the value of DR is calculated. The element compositions of the cladding layer in different areas are analyzed. The microhardness is measured. The results show that the variation trends of DR obtained by numerical simulation and experiment are basically consistent, and DR decreases gradually with the increase of scanning speed. The higher the scanning speed, the smaller the dilution effect of elements in the cladding layer is. The nearer the spot to the substrate, the more obvious the dilution effect of element composition in the cladding layer is. The microhardness from cladding layer to substrate shows the three-stage distribution of low-high-low. The width of the high hardness zone decreases gradually with the increase of scanning speed.

A kind of adaptive optical principle correction system based on compound liquid crystal spatial light modulator (LC-SLM) is designed and built. One spatial light modulator is used for wavefront simulation and the other for wavefront correction. The Zygo interferometer is used to detect the distorted wavefront. The conjugate wavefront of the distorted wavefront is obtained by the program calculation, and the corresponding gray image is generated, which is loaded into the LC-SLM to compensate the distortion wavefront, and the static wavefront correction is realized. The calculated Stell-ratio increases from 0.3795 before correction to 0.8268, and the mean value of the ring girth gray increases from 382.75 before correction to 1164.5, and the luminance aggregation of the spot is obviously improved. Compared with the pure phase space light modulator, the composite space light modulator saves costs and is suitable for laboratory batch use.

The temperature and stress fields of the laser cladding are numerically simulated by ANSYS birth-death element technique based on the thermo-mechanical indirect coupling nonlinear finite element analysis. The results show that this proposed model is reliable by analyzing the temperature distribution of the finite element model and the morphological characteristics of the metallographic structure of the experimental specimens. The temperature change of the cladding layer can be divided into two stages. In the first stage, the temperature rises rapidly like a pulse. In the second stage, the temperature drops to the overall temperature of the matrix which likes a hyperbolic shape. On the surface of the cladding layer and along the laser scanning direction, the peaks of the temperature time curves of multiple nodes show a trend of gradual increase. The distribution curves of the residual stress show that there is a greater residual stress at the position near the fixed end along the Z axis in the middle of the bonding surface of the cladding layer and the matrix. The residual stress is distributed symmetrically like the shape of W along the X axis in the middle of the matrix undersurface which is likely to cause stress concentration and mutation at the interface between the cladding layer and the matrix along the Y axis in the middle of the free end.

An optimization method for measuring network node pose is proposed. The mathematical model of network node parameters is established, the influence mechanism of pose parameters on the measurement accuracy is analyzed, and the optimal range and optimal initial solution of pose parameters are obtained by simulation. Distance between the mark points on calibration plate is measured by three-dimensional vision. Experimental results show that the relative error of measurement after the optimization is reduced by 0.6%.

Aiming at the difficulty of binocular synchronous stereo matching under high-speed camera conditions, a fast measurement algorithm for the spatial linear motion trajectory based on binocular camera is proposed. The camera space coordinate system is established ,and the imaging model is constructed based on the principle of small hole imaging and the mapping relationship between the target point and the corresponding point. The principle of the algorithm is verified experimentally, and the method of reducing the error by traversing the input points and finding the average value of the result is obtained. The feasibility of the principle and the effectiveness of the algorithm are verified.

An optical model of Chebyshev-shaped particle with ellipsoidal nucleus is established for agranulocyte, aiming at their morphological changes due to the disturbance of extracellular fluid. Based on the T-matrix method, the scattering characteristics of linear polarization of the white blood cells (WBCs) at different morphometric parameters and nucleus to radius ratios are numerically simulated for linearly polarized incident light. The scattering polarization characteristics are compared with those of WBCs without the disturbance of extracellular fluid. The research results show that the degree of linear polarization of the scattered light is more sensitive to the external liquid disturbance at the side-scattering region than that at the backscattering region, the stronger the external fluid disturbance, the greater the influence on light polarization. Moreover, the larger the nuclear deformation parameter, the greater the influence of extracellular fluid on the polarization of scattered light.

The transfer matrix method is used to study the optical transport process in a terahertz frequency region in grapheme-containing multilayer dielectric structures. In the multilayer structures, the optical absorption exhibits as multiple absorption peaks, and the position of these peaks depends on the incident frequency, incident angle and graphene chemical potential. With the increase of the graphene chemical potential and the layer number of the graphene, the absorptivity increases and the imaginary part of the graphene conductivity increases, and the position of the absorption peak shifts blue. The periodic absorption peak occurs under the condition that the reflected light interference is cancelled, and can be understood by the phase difference between the two main reflected lights. Therefore, the chemical potential and energy level broadening factor parameters of the graphene can be given based on the position and intensity of the graphene absorption peak in the light transmission. The parameter conditions related to the graphene materials which can improve the absorptivity of graphene and reduce the transmittance are given.

A scheme for asymmetric manipulation of Rayleigh particles using elliptical vector hollow beams is presented. The scattering force and the gradient force of the elliptical vector hollow beams on the Rayleigh particle which are in these beams are theoretically calculated. Taking the elliptical vector hollow beams produced in the experiment as the manipulating beams, and taking the water molecules group in acetyl benzene liquid as the research object, the theoretical analysis of the trapping dynamics of the related particles is carried out, and the Monte-Carlo method is used to simulate the trapping dynamics. The theoretical research results show that the scattering force onto the Rayleigh particles in elliptical vector hollow beams is much smaller than that of the gradient forces. The Rayleigh particle can be trapped in the asymmetrical hollow part of the beam by choosing the appropriate light intensity.

A method for cloud detection of ZY-3 satellite remote sensing images is proposed based on improved deep learning fully convolutional neural network. In pre-trained deep convolutional neural network, full convolution layer is used instead of full connection layer, and deconvolution method is used to up-sample feature map to optimize and improve network structure, then the Adam gradient descent method is adopted to accelerate convergence. The network is trained by using the resource image database of ZY-3 satellite, and the up-sampled image features are input into the Sigmoid classifier . Experimental results show that the proposed method performs better than the traditional methods in terms of detection accuracy and speed. The accuracy achieves 90.11%, and detection time can be reduced to 0.46 s.

A calculation method of complex tank capacity is proposed based on the laser point cloud and building information modeling (BIM) technology. The bulkheads and inner components are first segmented based on the three-dimensional (3D) point cloud data of complex tank. The 3D reconstruction is completed based on the point cloud data of bulkheads and inner components through the BIM technology. The real-time calculation of tank capacity in different attitudes is done through the 3D model of tank. The proposed method is evaluated through the point cloud data of a complex tank in Shanghai. The research results show that this method can be used to quickly and accurately acquire tank capacities in different attitudes with high reliability and high practical application value.

Based on the principle of air refractive gradient measurement, a non-contact flame temperature field measurement instrument is designed using the projective background-oriented schlieren technique. A laser diode is used as the light source. Based on the CCD fast imaging and the particle image velocimetry technique, the offset of particles in the photograph is obtained to quantify the deflection angle. The refractive index gradient is obtained by Radon transform. The temperature field in each region of the flow field is obtained directly using the nonlinear curve fitting equation of air refractive index to temperature. The three-dimensional non-uniform temperature field is reconstructed by the back-projection algorithm. Thus, the visualization of flame temperature measurement is achieved.

For the terrain classification problem of light detection and ranging (LiDAR) point cloud data in complex scenes, a deep neural network model based on multi-scale features and PointNet is proposed. The method improves the ability of PointNet to extract local features and realizes the automatic classification of LiDAR point cloud under complex scenes. Multi-scale network on the basis of PointNet network is added to extract the local features of points, and the local features of different scale points are formed into a multi-dimensional feature through the full connection layer, and combined with the global features extracted by PointNet, the score of each point class is returned to complete the point cloud classification label. The proposed deep neural network model is verified by using the Semantic three-dimensional dataset and the Vaihingen dataset provided by ISPRS. The research results show that the proposed algorithm achieves higher classification accuracy compared with other neural networks for point cloud classification.

When the polymethyl methacrylate (PMMA) layer is placed between the gold nanoparticles and gold substrate, we study the effects of metal nanoparticles on the total emission rates and radiative emission rates and far-field radiation pattern of the radiation source of molecular or quantum dots in the PMMA layer. Compared with single gold nanoparticle, when two, three, five and nine gold nanoparticles are put on the PMMA layer with the same distance, the total radiation rates and radiative emission rates are enhanced. By measuring the far-field radiation pattern, the number of nanoparticles and the distance between nanoparticles can be inversely deduced, thus realizing the high precision measurement of nanoparticles.