A vortex beam can effectively improve the channel capacity of the communication system, however in the atmospheric environment, the optical communication channel is influenced obviously by the atmospheric turbulence. Thus it is of great significance to study the transmission characteristics of a vortex beam in the atmospheric turbulence. The Kolmogorov spectral model is widely used to describe the atmospheric turbulence previously, but the further researches show that the atmospheric turbulence also has the non-Kolmogorov spectral characteristics. Therefore, the investigation of the non-Kolmogorov spectral transmission characteristics for a vortex beam is conducted. Based on the numerical methods, the influence of outer- and inner-scale of turbulence, generalized exponential factor, refractive-index structure constant on the spiral spectral distributions, topological detection probabilities and others of a Laguerre-Gaussian beam after different transmission distances is investigated. The simulation results show that the topological detection probability is closely related with the above parameters. Finally, a numerical method for calculating the scintillation index is proposed, and the influence of turbulence on bit-error-rate (BER) is figured out. The results show that when the beam is propagated up to 1000 m even in a weak turbulence, the BER is still difficult to meet the communication requirements, and thus a further phase correction is quite necessary.

Due to the limited network nodes in the optical network link, the multiple wavelengths simultaneously inputted in network compete for the same output port and thus the communication network is made obstructed. Thus a Te-based Raman wavelength converter is applied to the 20×20 Mesh-Torus network when its placement rate is 50%, and its influence on the network blocking rate performance is analyzed. With the Bernoulli-Poisson-Pascal(BPP)network model, the effects of input wavelength number S and degree of conversion d of the Te-based wavelength converter on the blocking performances of the Mesh-Torus network are simulated and analyzed. The results show that when d is fixed, the variance of blocking rate is within 0.088-0.0945 for S=4 and within 0.0960-0.0995 for S=6, indicating that the increase of S slows down the blocking of network communication and the blocking rate increases slowly. In contrast, when S is fixed and with the increase of d, the blocking rate decreases approximately by 0.0005-0.004 on the original basis. Therefore, it is feasible to reduce the blocking rate of the Mesh-Torus network with a Te-based wavelength converter.

In ultraviolet (UV) scattering communication systems, pulse broadening and noise considerably affect the signal. To reduce system bit error rate and increase communication distance, it is necessary to add low-density parity check code (LDPC) to implement symbol error correction. The performance of on-off keying (OOK) modulation and binary phase-shift keying (BPSK) for subcarrier intensity modulation is compared. Further, the influence of LDPC codes with different coding efficiencies and decoding methods under BPSK subcarrier intensity modulation on the bit error rate of a non-line-of-sight (NLOS) ultraviolet communication system is simulated and analyzed. Results indicate that BPSK subcarrier intensity modulation performs better than OOK modulation. LDPC codes with low coding efficiency values exhibit strong error correction ability. The performances of belief propagation (BP) algorithms and log-likelihood ratio belief propagation (LLR BP) algorithms are virtually the same, and both are better than min-sum algorithms. When the bit error rate is 6.5 × 10 -6, the communication distance of the LDPC (672,168) coded sequence is approximately doubled, compared with the uncoded sequence.

Bio-cell imaging is one of the important research directions in the biological and physical fields, especially the measurement and analysis of large amount of cells have major research value and significance in the bio-research and disease-diagnosis. However, the traditional microscopic detection is only restricted to the qualitative observation of cellular contours. Moreover, the amount of measurement samples is difficult to satisfy the statistical requirements due to the limitation of field of view. Thus a quantitative interferometric microscopic cytometer is designed based on the combination of quantitative interferometric microscopy and field of view scanning, and is used for the imaging and analysis of large amount of cells, in which the quantitative interferometric microscopy is capable of retrieving the cellular phase distribution, and of obtaining and analyzing a large amount of cellular information by combining the field of view scanning. The expanded principle component analysis algorithm, regularized optical flowing algorithm and Fourier phase retrieval algorithm are used, respectively. In addition, the different scanning modes are also combined, and thus the measurements and statistics of phase area, phase volume, circle ratio and other parameters of cells are obtained. The whole system is expected to obtain applications in high-throughput and high-speed cellular detection.

The super-resolution algorithm based on the convolutional neural network has great advantages compared with the traditional super-resolution algorithms. But there are still some problems to be improved, such as long training time, lacking of image texture reconstruction and so on. Owing to this, on the basis of the original convolutional neural network super-resolution reconstruction algorithm, the following optimizations are carried out. The original rectified linear unit function is discarded and the new activation function is used instead. The network structure is changed and image reconstruction is achieved by the final deconvolution upsampling. The original stochastic gradient descent optimization algorithm is replaced by adaptive moment estimation algorithm whose optimizes performance is faster and better. Comparative experiments are carried out on Set5 and Set14 test sets, respectively. The experimental results show that the reconstruction effects of the improved method with less training time are greatly improved on the objective evaluation index, for example, the power signal-to-noise ratio increases up to 2.33 dB, and the texture is clearer, the edges are more complete and the reconstruction effect is better on the subjective visual effects.

An algorithm of image salient object detection combined with deep learning is proposed based on an improved recurrent deep convolutional neural network with the cross-level feature fusion. The feature extraction of input images is performed through this improved recurrent deep convolutional neural network model. The cross-level joint framework is used for the feature fusion and thus the initial salient maps with high-level semantics features are generated. The saliency propagation is applied to the fusion of initial salient maps and low-level image features, and thus the structural information is obtained. The saliency propagation results are further optimized with the conditional random field and the final salient maps are realized. With the massive datasets, the proposed algorithm is tested and compared with other algorithms. The research results show that the proposed method is more robust than the existing algorithms in the image salient object detection of the complex scenes. Moreover, the integrity of the significant target detection is improved and the background is suppressed effectively.

According to the problem of the severe detail loss of the disparity map generated by the current convolutional neural network methods, a structural improvement method is proposed. The 4 layers convolutional structure of the feature extraction part from original network is added to 7 layers to maximize the accuracy. And, the proposed dual pyramid structure is introduced to the network to combine the multi-scale down-sampling information with the feature information, which keeps the details of the original input images. Experimental results show that the error rate of the improved network reduces from 3.029% to 2.795%, and the generated disparity maps have better connectivity.

Owing to the problems of the absorbing Markov random walk method failing to fully suppress the central background area of the saliency map and losing parts of salient objects near the image boundary, an image saliency detection method based on manifold regularized random walk is proposed. First, a global graph with superpixels from the input image is constructed. An initial saliency map is obtained by using the absorbing Markov chain, and then an adaptive threshold is used to segment the initial saliency map to get robust foreground queries. Second, in order to make effective use of the complementarity of global information and local information, an optimal affinity matrix is obtained by constructing the local regular graph. Finally, the obtained optimal affinity matrix and foreground queries are applied in the manifold regularized framework to obtain the final saliency results. Experimental verifications are carried out on three public datasets. The results show that the precision and recall rate of saliency detection have been improved by the proposed method.

In recent years, the single image super-resolution methods based on deep learning have made remarkable achievements. However, these methods focus researches on the image spatial domain, ignoring the significance of information in high frequency in image frequent domain, resulting in a relatively smooth image. A single image super-resolution method combining wavelet transform with deep network is proposed, which takes the advantages of extracting the details by wavelet transform. First, the image is decomposed into a sub-band in low frequency and three sub-bands of different directions in high frequency by wavelet transform, then the low resolution image and sub-bands in high frequency are regarded as the input of the deep network. Second, the existing deep network is improved by simplifying the network, decreasing the number of convolution layers to reduce network burden, and modifying the network channels. Finally, the super-resolution image is obtained by inversely wavelet transforming. The proposed method is tested on the open test datasets, and compared with other state-of-the-art methods. The experimental results demonstrate that the proposed method works well in subjective visual effects and objective evaluation indexes.

Aiming at the face images in large angle oblique illumination and extremely dark and uneven illumination environment, an illumination compensation method is proposed based on anisotropic Retinex transform. First, according to the statistical characteristics of face image, the direction of the light source is analyzed, and the edge is detected by Prewitt operator. Combined with the geometric characteristics of the face texture, the curvature, slope and symmetry are introduced to achieve the unevenness of the face and illumination, thus distinguishing the false edge of the face. Second, based on the Weickert structure tensor, an improved anisotropic diffusion model is implemented based on different types of edges. The model is combined with Retinex algorithm to realize face image illumination compensation. The experimental results show that the improved anisotropic diffusion method can enhance the image brightness, prominent texture detail, and eliminate most light shadow at the same time enhancing face edge.

For the fixed infrared camera video surveillance, and the poor imaging results and limited monitoring range in the night and foggy weather conditions, we propose the zoom color video construction method in night and foggy days based on LED infrared camera with PTZ, combining with the characteristics of infrared imaging and visible light imaging. This method uses visible light imaging to achieve panoramic image stitching during daytime with good weather conditions, then the infrared image with the panoramic image is registered to construct a visible light background image with the same field angle. An infrared imaging system is used to extract moving objects in night and foggy days with low visibility, then through the mask fusion algorithm and the registration parameters, the moving target is fused to the same position of the visible light background image in the same proportion, and finally the zoom color video is constructed. The experimental results show that this method can meet the requirements of wide monitoring, real-time monitoring of color video in night and foggy all days.

In the actual traffic environment, the quality of the collected traffic signs is often influenced by the factors such as motion blur, background interference, weather conditions and shooting angles and so on, which poses a great challenge to the accuracy, real-time and robustness of traffic sign automatic identification. Owing to this, a classification recognition algorithm model of improved deep convolution neural network AlexNet is proposed. On the basis of the traditional AlexNet model, this model takes the traffic sign image data set GTSRB taken in the real scene as the research object, modifies the convolution kernels of all coiling layers to 3×3, in order to prevent and reduce the occurrence of over fitting, the dropout layer is added after two fully connected layers. In order to improve the accuracy of traffic sign recognition, two convolution layers are added after the fifth layer of the network model. The experimental results show that the improved AlexNet model is advanced and robust in traffic sign recognition.

The problem of lossless compression of bi-level speckle images is studied with the statistical properties of dynamic speckle. A bi-level speckle image distortion coder is designed, which consists of image cutting, displacement estimation, prediction, context formation and entropy coding. The main features include the fast algorithm of speckle displacement estimation, the optimized prediction based on the statistical characteristics of speckle, and the adaptive arithmetic coding. The working process and principle of the coder are introduced. The experimental results show that the compression performance of the bi-level speckle image distortion-free encoder is greatly improved.

With the rapid development of three-dimensional (3D) scanning technique, the huge volume of point cloud has been produced, which puts forward higher requirements for the performance of point cloud computing. Therefore, how to improve the efficiency of the algorithm has become a hot topic in this field. There are rich 3D shape models hidden in the ever-increasing amount of point cloud data. Inspired by the relationship between 3D shape models and the point cloud, we provide a new method to improve the execution efficiency of algorithms about point cloud computing. The 3D geometric feature analysis technology is used to obtain shape-related feature parameters, and the point cloud segmentation algorithm is proposed based on it. We use octree algorithm to organize point cloud and obtain the neighbor relationship. A self adaptive and dual linear octree algorithm is designed based on the density of point clouds to establish the data index. We build a 3D shape library by using regular shape models, and realize the algorithm for matching models with data segmentation regions. Further, we extract the shape parameters of the segmented region, which are the foundation for improving the accuracy and speed of point cloud data processing. Moreover, the segmentation effectiveness and time performance of different algorithms are compared, and the experimental results indicate that the proposed algorithm is feasible and robust.

In order to remove the thick cloud of aerial images, an improved image restoration algorithm based on improved Criminisi algorithm is proposed. The optimal inpainted block is selected by improving the priority function, and the searching strategy of the best matching block is optimized. In this way, the structural propagation error and cumulative error are reduced, and the matching accuracy is improved. Then the suitable size of the sample block is selected according to the local variance of the brightness of the pixels. A new confidence updating function is defined. The thick cloud area mask can be obtained through gradually refined cloud detection method and morphological closed operation. The experimental results of simulated data and real data show that compared with the traditional Criminisi algorithm, the completion effect of proposed method is more natural, and the thick cloud area in aerial images can be better repaired.

A depth image stitching method based on binocular vision is proposed, in which a pre-calibrated binocular depth sensor is used to acquire two depth maps with certain overlapping areas through motion. These depth maps correspond to the pixel points of left-view images one by one. The feature extraction is performed on the left-view images and after matching, the homography matrix is calculated and obtained. Then the two depth maps are further stitched and the depth of these maps is corrected based on the homography matrix, and thus the final depth map stitching results are obtained. The simulation and experimental results show that, the proposed method can effectively expand the field of view of the binocular depth sensor and the obtained stitching depth map is basically same with the depth map acquired by a single sensor.

Aiming at the problem that the conventional lane detection system uses a single-channel forward-looking camera under night scenes, which is susceptible to strong light interference and is prone to false detection and misdetection in complex scenes, we propose a lane detection method based on active infrared filter and around-view imaging. In the imaging stage, four-way vehicle-borne cameras based on active infrared filter are used to collect scene information around the vehicle, and then a look-around image with 360° overlooking effect is obtained based on perspective transformation and image fusion. In the detection phase of lane, a lane detection algorithm is proposed based on agglomerative hierarchical clustering. Firstly, based on the shape features of lane lines, a more pertinent template matching is designed to extract the edge points of the lane line. Then the edge points are clustered by agglomerative hierarchical clustering, and the lane is fitted by the random sample consensus algorithm. Finally, a priori information and Kalman filter are combined to further improve detection accuracy. The results show that the proposed algorithm can effectively eliminate the strong light effects during the detection of lanes and effectively reduce the false detection and missed detection rate to a certain extent.

A target tracking algorithm via confidence evaluation is proposed to add the scale and model updating method to the correlation filter. In the process of tracking, there are many cases of occlusion and similar interference. If the model parameter is continuously updated, it can easily lead to false track and target loss. Therefore, the quality of tracking is judged qualitatively by the confidence. When the confidence is low, we stop updating to prevent the introduction of error and improve the accuracy. After ensuring the tracking is correct, we can detect and update the scale size. We propose a faster scale updating method with redundant code simplified, and make the tracking more accurate and with lower time cost. The experimental results show that the proposed algorithm improves the precision and success rate by 38% and 33%, respectively compared to the original algorithm, it has better performance than several existing algorithms, and is more robust to cope with occlusion and scaling.

In order to solve the infrared image registration problem of the monitored exhaust in vehicle exhaust monitoring system, a method of infrared image registration based on double scale search genetic algorithm is proposed. The method takes mutual information as the similarity measure and modify probability formula of crossover and mutation in the traditional genetic algorithm. The proposed double scale search genetic algorithm is used as the optimization algorithm to realize high precision registration of vehicle exhaust infrared image. The obtained root-mean-square error of horizontal translation, vertical translation and rotation angle are 0.0949, 0.0447 and 0.0000, respectively, and the results of experiment with this method are better than other methods and it proves the effectiveness of the proposed method. In comparison to the image registration method based on the adaptive genetic algorithm and ant colony algorithm, and the proposed method has higher precision and better stability. Compared with the Powell algorithm, the proposed method has stronger anti-noise ability and is more suitable for exhaust image registration, which is a good foundation for inversion and calculation of pollution gas concentration.

The flow cytometry nowadays includes non-imaging-type and imaging-type. The non-imaging flow cytometry has a high throughput, but it cannot provide cell morphological information. The imaging flow cytometry which can obtain cell morphological information has a low throughput due to the frame rate limits of the conventional CCD array imaging sensors. Aiming at this problem, an ultrafast flow cytometry quantitative phase imaging system based on optical coherent detection is proposed and designed in this paper. Its effective line scan imaging speed is up to 100 MHz. The cell imaging experiment results show that the proposed system can acquire high-contrast quantitative cell phase images and a theoretical throughput of 50000 cells per second.

In order to improve the detection and recognition ability of an electro-optical imaging system, a novel method for the target-background contrast enhancement based on color polarization imaging is proposed. Based on the Fresnel principle, the variation of target degree of linear polarization with wavelength of incident light is investigated. A collection of color polarization images under different rotation angles is conducted, which is separated into 0°, 45°, 90° and 135° polarization images within the red, green and blue channels. In addition, the polarization images within each channel are calculated, respectively. The red, green and blue polarization images from three channels are synthesized as a novel color polarization image, which contains richer image information than its original color image. A color polarization experimental setup is constructed. Two groups of experiments on target-background contrast enhancement are conducted, which take vector angle distance and contrast calculation as two objective evaluation indexes. The results show that this color polarization method can be used to increase the average distance of vector angle by 0.13 and the contrast calculation by 0.11.

The single-frequency tunable tapered semiconductor laser amplification system with maximum output power of 1.3 W and wavelength of 910-930 nm is reported, and variations in the output power of tapered semiconductor amplifier with the injection current and seeding power are experimentally investigated. Through the isolator and focusing lens, the 13.6 mW single-frequency seeding light at 920 nm decreases to a power of 12.4 mW. The output power of the amplifier with injection current of 4 A can reach 1300 mW, the gain is up to 20.21 dB, and the linewidth is 660 fm. Moreover, when the seeding power increases from 0 to 13.6 mW, the amplification power increases accordingly. The laser obtained by this system can be used to study the narrow linewidth continuously tunable medium (deep) ultraviolet laser after quadrupling.

This study investigates the morphological modification of gold nanoparticles using ultrashort laser pulses and subsequent changes in the resonance spectra of localized surface plasmon polaritons (LSPPs) known as phase transition. A thin film comprising a random array of gold nanoparticles is prepared using chemically synthesized colloidal gold nanoparticles. When the femtosecond laser pulse is irradiated on the film surface, changes are observed in the microscopic image and the extinction spectrum before and after laser irradiation. The measurement results show that the femtosecond laser pulses induce the melting and aggregation of the gold nanoparticles in some microscopic areas, leading to a morphological change of the gold nanoparticles and a red-shift in the plasmonic resonance spectra.

Based on the single-periodic and multi-periodic oscillation states of an optically injected semiconductor laser, a scheme for the generation of an ultra-broadband microwave frequency comb through the cascaded optical injection of semiconductor lasers is proposed. In this scheme, the optical frequency comb produced from a slave laser with continuous-wave injection is injected into another one, to generate a microwave frequency comb with much wider bandwidth due to the optical injection effect of laser. The numerical investigation based on the rate equations of an optically injected semiconductor laser shows that, as for the second optical injection and the selection of appropriate parameters, the bandwidths of this microwave frequency comb reach 52, 65 and 97 GHz within amplitude fluctuations of ±2.5 dB, ±5 dB and ±10 dB, respectively. Therefore, the proposed cascaded injection scheme is effective in generating ultra-broadband microwave frequency combs with more flat property.

A Co∶MgAl2O4 crystal with an initial transmissivity of 89.5% is used as a saturable absorber at a wavelength of 1.3 μm. A low-threshold and high-efficiency passively Q-switched 1319 nm Nd∶YAG laser output is achieved with two-mirror resonator. When the pump power is 10.0 W, a pulsed laser with an average output power of 490 mW, the narrowest pulse width of 18.3 ns, a repetition rate of 17.5 kHz, a pulse energy of 28.1 μJ and a peak power of 1533 W is obtained. The results show that the Co∶MgAl2O4 crytal possesses a good saturable absorption property at a wavelength of 1.3 μm.

The green parts of cordierite/carbon-fiber composites are fabricated by the selective laser sintering technique and then the final porous ceramic parts are obtained by high-temperature sintering. The changes of porosity, mechanical performance and dimensional accuracy under different processes are investigated. The results show that the carbon fibers are bonded into the matrix by bonding necks and with the increase of sintering temperature, the cordierite melts and deforms, which leads to the reduction of porosity. When the sintering temperature is 1350-1400 ℃, the μ-cordierite is stable. In contrast, when the sintering temperature increases to 1425 ℃, the μ-cordierite is transformed into the α-cordierite. Moreover, the carbon fibers effectively enhance the toughness and strength of the green parts and with the increase of sintering temperature, the compressive strength of ceramic parts increases until to its maximum value of 5.48 MPa at 1425 ℃.

High energy nanosecond pulsed ytterbium-doped fiber laser is developed. The laser utilizes active Q-switching technology to generate nanosecond pulses, and the pulses can obtain high energy and average power after it passes through two-stage master oscillator power amplifier (MOPA) system. When the pulse repetition frequency is 100 kHz, the maximum average power of laser pulses is 302 W, pulse duration is 203 ns, single pulse energy is 3 mJ and the optical-to-optical conversion efficiency is 84%. Laser rust removal experiment is carried out using the laser, the experimental results show that the laser is stability, and the laser can meet the laser rust removal, laser paint removel and many other applications.

The laser can be influenced by the atmospheric turbulence during its transmission and this turbulence introduces aberrations which leading to the spot diffusion on the imaging surface and the decrease of energy utilization. Thus, the remote sound collection quality is reduced, which is not conductive to the application of laser eavesdropping in the actual environments. With aberration as the research object, the influence of aberration on the energy utilization in laser remote sound eavesdropping is investigated and the corresponding theoretical deviation is performed. In addition, the influences of single-order Zernike aberrations and the atmospheric turbulences with different types and different magnitudes on the remote sound collection quality are analyzed. Meanwhile, the numerical simulation is conducted. The results show that the influences of aberrations with different types and different magnitudes on energy utilization are different, and the utilization decreases with the increase of aberration. Moreover, the different types of atmospheric turbulences have different effects on energy utilization, and the utilization basically decreases with the increase of D/r0. This study provides a theoretical basis and reference for the improvement of laser remote sound eavesdropping quality.

In order to apply laser-induced breakdown spectroscopy (LIBS) into the monitoring and analysis of nuclear materials, we use pulsed laser with wavelength of 1064 nm to induce the plasma spectrum in the experiment, detect the optical signal using multi-channel grating spectrometer, and study the effects of different experimental parameters on the spectral property. The characteristic spectral lines Ce I 500.91 nm and Ce II 446.02 nm are chosen for analysis. The experimental results show that the experimental parameters such as the target position, the distance of the focusing lens to the samples, the excitation energy and the delay time have great influence on the LIBS signal of cerium oxide. High spectral intensity and signal-background ratio are obtained by the optimization of these experimental parameters, and the optimum experimental conditions for the analysis of cerium oxide by LIBS technology are determined, which provide some data reference for the future study of component analysis of plutonium oxide.

The mechanism of the laser-induced surface acoustic wave interacting with surface-breaking defects is studied by using the finite element method. The characteristic quantity (oscillation signal) quantifying defect depth is selected from the displacement signals and the influences of the defect depth and the width on the oscillation signals are further analyzed. The source of oscillation signals is clarified according to the displacement field occurred in the interaction between the surface acoustic wave and the rear edge of defects. The numerical results show that the oscillation signal with feature points of A and B originates from the oscillation induced at the rear edge of defects by the transmitted surface wave. At the same temperature, the arrival time difference between feature points A and B increases linearly with the defect depth, but is independent of the defect width. According to the relationship between the arrival time difference and the defect depth, the quantitative calculation of defect depths at different temperatures is realized by combining the relationship between the surface wave speed and the temperature.

Most tracking algorithms cannot solve the problem of scale variation and the existing scale solutions are redundant and fixed. To solve the problems, a fast and novel scale estimation method based on kernel correlation filtering framework is proposed, which is coarse-to-fine. Considering that the peak value of the response graph is not stable, the Euclidean distance of the detection response graph and the expected output graph are used as the reliability of the peak value. The product is taken as the final comparison result. Firstly, three scale factors are used to determine the direction of scale variation, and then solve the optimum in the direction of scale variation. The proposed algorithm is experimented on 26 benchmark sequences with scale variation attribute of OTB-100, and is quantitatively and qualitatively compared with other existing advanced tracking algorithms. The results show that the proposed method can solve the scale variation problem well. The proposed method is 18.8% higher in mean distance precision and 19.6% higher in area under curve than those of the kernel correlation filter. The tracking speed is 2.5 times of the accurate scale estimation for robust visual tracking, and is 6 times of the scale adaptive kernel correlation filter tracker with feature integration.

A method based on quantum-behaved particle swarm algorithm is proposed to optimize camera parameters. The intrinsic and extrinsic parameters of the camera are quickly obtained by the self-calibration program in MATLAB software, and the camera parameters are optimized by using the quantum-behaved particle swarm optimization algorithm. The experimental results show that the optimization algorithm can converge quickly and with high precision, and it can improve the camera calibration accuracy to some extent.

Estimating the number of people in the surveillance scene is one of the important tasks of security monitoring. However it is difficult to estimate the number when the crowd is with clutter and severe occlusion. An improved crowd counting method based on the convolution neural network is proposed as for the number estimation under dense scenes. In order to reduce the effect of camera perspective distortion, the deep network and shallow network are used to extract the crowd characteristics, respectively. The convolution layers with different kernel sizes are also designed. Moreover, the extracted features are fused through a special structure with multi-scale extraction capability. The experimental results show that the crowd density map obtained by the improved network model is closer to the original scene information and the obtained prediction results are more precise.

A smoke concentration grading method based on the image analysis technique is proposed for the automatic speed adjustment of the dust removal fan in the copper smelting process. We obtain a sequence of sub images by using a moving window to slide over the whole smoke image from top to bottom. Then, discrete cosine transform (DCT) is utilized to extract the features of each sub-image and the DCT coefficients are vectorized as the observation data for hidden Markov model (HMM). Thus an image is divided into an observed sequence to build the HMM model for grade classification. Four different running states are considered in the smelting process, in which a HMM model is built for each running state. For each running state, 30 images are used for the training of HMM model. The results show that the classification accuracy can reach 95% with HMM, which is higher than that of least squares support vector machine (LSSVM).

A face recognition algorithm based on the angular distance loss function and densely connected convolutional neural network is proposed under the open-set protocol to achieve deep face recognition. The loss function based on angle distance is adopted in the proposed network structure, which makes the facial features more distinguishable and meets the ideal criteria of feature classification. At the same time, the advanced dense connection module is adopted in the proposed neural network structure, which greatly reduces the parameter redundancy of the traditional network structure. After extensive analysis and repeated experiments, the face recognition accuracy reaches 99.45% on the LFW dataset, and the recognition accuracy rates of face identification task and face verification task on MegaFace dataset are 72.534% and 85.34%, respectively. The superiority of the proposed algorithm is confirmed in the face recognition domain.

A reconstruction method of three-dimensional texture mapping and imaging based on phase-assisted system is proposed, which can map the captured texture images around objects onto the surface of the geometric model to achieve the fusion of texture color of geometric surfaces. The texture mapping optimization function is established, and the mapping between the target object and the texture image is optimized. The texture fusion processing is used to ensure the fidelity of the target object texture to the greatest extent, and the three-dimensional texture reconstruction of the object is realized. The locally missing target model and the uncollected local texture are also repaired, and the natural transition of the three-dimensional texture surface is realized, thereby ensuring the integrity and consistency of the target model texture. The experimental results of different types of physical objects prove the feasibility and effectiveness of the proposed algorithm.

In this paper, a planar and vertical combination of metamaterial that can realize the multiband and broadband electromagnetically induced transparency effect by three-dimensional coupling is designed. Through the coupling of three vertical split ring resonators(SRRs) and one planar square closed loop(SCL), the electromagnetically induced transparency of the metamaterial is realized at 0.68 THz and 1.09 THz. The bandwidth can reach 0.38 THz and 0.74 THz, respectively. By comparing and splitting the structures, the physical mechanism of electromagnetically inducing transparency through multiband and broadband is studied. Besides, the influences of the distance between the three SRRs and their arm-lengths on the intensity and bandwidth of the electromagnetically induced transparency are analyzed. The simulation show that the structure of metamaterial can achieve slow light effect with high intensity at multiple frequency points and high refractive index sensitivity in terahertz range. It has some application value in the field of optical buffer devices and refractive index sensing.

A high resolution wide-spectral micro-spectrophotometer is developed based on the asymmetrical cross Czerny-Turner system. The wavelength of this spectrophotometer is calibrated by the three polynomial fitting method, and the relative error of wavelength calibration is less than 0.1 nm. This spectrophotometer is used to set up an experimental measurement system, and the chromogenic agent and the standard chromium solution are formulated to test the performances of this system. The results show that this spectrophotometer can meet the application requirements of the online monitoring of hexavalent chromium water quality, which provides a solution for realizing high-precision, multi-wavelength water quality parameter online monitoring.

An algorithm used to design a free-form lens for the extended light-emitting diode (LED) source is proposed. The design of a free-form lens is realized by the overlap of the weighted lens profiles of several point sources and the evaluation function reflecting the illumination uniformity is constructed, in which the independent variables are three weighting factors. The particle swarm algorithm is used to optimize the weight factors and minimize the evaluation function, and thus the illumination uniformity is the highest. Correspondingly, the illumination uniformity on the target surface increases from 78.93% for the initial structure to 90.67% and simultaneously the light efficiency is kept above 92%. The fabrication error analysis and lateral tolerance analysis of lens show that the influence on the illumination uniformity on the target surface is small when the fabrication error is within the range of -1.0-1.0 μm and the lateral displacement error is within the range of 0.0-0.2 mm.

A projectile spotlight is designed based on the light emitting diode(LED) source with a high power and a high optical density. Based on the non-imaging optical theory and by using the ZEMAX software, a collimating optical system with the large-aperture and high-transmissivity Fresnel lens as the core element is established. A cylindrical Fresnel lens with a focal length of 1500 mm is used for the diffusion of the collimated beam along one-dimensional direction and then the spotlight is irradiated at the target surface with a tilt angle of 6.85° and finally a large area illumination is realized. The simulation analysis by the TracePro software shows that the divergence angles of light spot in the horizontal and vertical directions are 18.1°and 2.46°, respectively. A clear and high contrast oval-like shaped bright spot is obtained when the prototype is projecting on the aiming target, and the maximum illuminance at its center is over 110 lx, which meets the requirement of the night shooting for fighters.

Optical zoom system based on liquid lens is used in many optical fields in recent years because it can achieve zooming without mechanical movement. A new lens with three layers of liquids and connected conducting layer is proposed. This liquid lens is cylindrical and it can zoom and compensate by itself by changing the voltages applied to control the radium of liquid interfaces. The structure of the liquid lens is analyzed in detail, and the minimum structure expression of the liquid lens is obtained. On the basis of the Gaussian optics theory, the effects of lens parameters on zooming performance including zoom ratio and field angle of the system are studied.

Metal nanomaterials are widely used in semiconductor materials luminescence, solar cells, surface enhanced Raman scattering detection, and photochemistry, etc., because of their unique local surface plasmon resonance (LSPR) characteristics. Owing to its very low absorption loss in a particular band, Ag is considered as an excellent candidate material for LSPR. In this paper, the near-field local enhancement and far-field scattering characteristics of cylindrical Ag nanostructures are systematically simulated and analyzed by finite-difference time-domain method (FDTD). The results show that the size, spacing and substrate refractive index of Ag nanostructures have significant effects on the LSPR. Furthermore, the LSPR properties of Ag nanostructures can be regulated by changing of the structural parameters.

With the recently increasing influence of haze/fog and other inclement weathers and the expanding impact of smoke and dust in the battlefield on military operations, the identification, detection and recognition of targets in smoke and fog environment draw much more attentions. Thus based on this point, the polarization detection defogging technology based on multi-wavelet fusion is proposed. This technology is built on the target polarization detection and simultaneously combined with the merits of multi-wavelet, such as symmetry, orthogonality and compact support. According to the features of the high and low frequency coefficients, the different fusion rules are adopted for the fusion of target polarization information. This algorithm is much easier to get the prominent target contour and detail information than the traditional wavelet fusion. A large number of outfield target detection experiments are applied in the actual fog and haze environment. Both the subjective and objective evaluation criteria show that this technology has obvious advantages in the subjective visual effect and the objective evaluation aspects of contrast, definition, spatial frequency and so on, which can increase the identification efficiency of targets in fog environment.

A measurement method of Müller matrix is proposed based on the trajectories on the Poincare sphere. A matrix obtained by the polar decomposition is related to its trajectories on the Poincare sphere. Therefore, the birefringence and dichroism matrices of a nondepolarizing unit are directly obtained via the position relationship of the input and output light polarization states on the Poincare sphere. The results show that, based on the proposed method, only one rotatable polarizer is required in the polarization control system. With the measurement of two polarization states determined before and after tests, not only the birefringence and dichroism matrices can be directly obtained, but also the state of the fiber to detect does not need change during the measuring process.

Most of the existing anomaly detection algorithms for hyperspectral image only focus on the spectrum differences between the target and background while ignoring the spatial structure differences, which leads to poor detection results. Aiming at this issue, we propose a novel algorithm based on the combination of spectral and spatial information for anomaly detection (SSAD). To preserve the spatial structure information of the image, we detect anomalies band by band. The dual windows are established to calculate the luminance differences between the pixel under test (PUT) and background, and the spectral anomaly degree of PUT is measured. Then the inner window is regarded as the spatial structure window of PUT, and the most similar spatial structure window with the spatial structure window of PUT is searched from the background. The differences between the two is calculated to measure the spatial structure anomaly degree of PUT. Thus, the anomaly index of the PUT is obtained by the measurement of spectral and spatial anomaly degree. Going through the whole image, the detection result of the algorithm is acquired by summing up the anomaly index of each band correspondingly. Experimental results on three hyperspectral data show that, compared with existing anomaly detection algorithms, the proposed algorithm can significantly reduce the false alarm rate and has good robust to noise.

Extracting individual tree crowns in high resolution remote sensing image can improve forest inventory and management. To solve the problem that the existing individual tree crown extraction method has low accuracy in broad-leaved forest with high tree crown density, we propose an iterative H-minima watershed method for individual tree crown extraction in high resolution remote sensing images. Firstly, the morphological open operation is used to smooth the image, the Sobel operator is used to extract the gradient image, and the mean filter is used to denoise. Secondly, a set of h values are iteratively used to identify tree crown markers on gradient images, and the invalid markers are filtered by using the false marker detection method. Finally, the symmetry principle is introduced to restrict the flooding process of watershed algorithm, thus avoid the overgrowth of tree crown and combination of unmarked tree crowns. The high resolution remote sensing image is used as the data source, and the traditional marker-controlled watershed algorithm and the proposed algorithm in this paper are used to extract the single tree crown. The quality of the individual tree crown extraction is evaluated according to the single tree position and crown contours, and from both samples and individual tree scales. The results show that the F measurement of the tree crown obtained by the proposed algorithm in this paper is 92.71%, which is 31.99% higher than that of the marker-controlled watershed algorithm. This proposed algorithm can effectively suppress the over-segmentation, reduce the under-segmentation, and improve the extraction precision of the individual tree crown.

Cell lasers is a frontier interdisciplinary research field of laser photonics and life sciences, the principle of which is that, under the optical feedback of the cavity, the weak signal can be oscillated and amplified via the combination of fluorescent protein, bio-compatible fluorescent dye and luciferin into cells in the fluidic environment. We present a detailed discussion about the research status and basic principles of Fabry-Perot cavity and whispering-gallery-mode microcavity. Laser-based detection can effectively enhance the sensitivity and resolution compared with the tradition fluorescence-based detection, benefitting from the feedback amplification of cavity. The physiological changes inside cells are studied with the analysis of the spectra and modes of the cell laser emission simultaneously. It'll provide new technologies and design ideas for medical diagnosis, three-dimensional super-resolution imaging of biomaterials and integrated light source research.

The photonic lantern is an optoelectronic device emerging in recent years. It provides low-loss interfaces between single mode and multimode systems, with one side of the device being a multimode waveguide satisfying special designs, and the other side being a bunch of single mode waveguide. The theory and several typical structures are analyzed, and their characteristics and applications are summarized. The applications and research progress of the photonic lantern in astrophotonics, multiplexing, mode control, and high power laser are introduced, and its applications as a low-loss optical waveguide device in future optical systems is forecasted.

The femtosecond optical frequency comb is becoming an increasingly valuable research field in laser optics. In recent years, all polarization-maintaining (PM) fiber-based lasers and optical frequency combs have achieved rapid growth owing to the continuous improvement of both PM-fiber and device fabrication technology. Herein, the development of all PM-fiber-based optical frequency combs is reviewed from a technical perspective. First, the basic principle, framework, and key components of an optical frequency comb are briefly introduced, and each key technology of the frequency comb, such as self-started mode-locking, is discussed. Second, several methods for pulse amplification and pulse compression are described, such as chirped-pulse amplification, nonlinear amplification, and divided-pulse amplification. Third, supercontinuum generation and self-referenced interferometry technology related to carrier phase offset(f0) locking are introduced, and the f0 signal with a signal to noise ratio (SNR) as high as 40 dB is experimentally demonstrated. Finally, the methods for locked repetition frequencies(fr) and f0 signals are illustrated.

As an important renewable energy technology, concentrating solar power (CSP) plays an important role in the development and utilization of solar resources, supply of energy, energy conservation, and emission reduction. The surface energy flux distribution of the receivers of four CSP systems including parabolic trough system, linear Fresnel reflector system, solar power tower, and solar dish system are analyzed and reviewed. The challenges brought by the non-uniform flux distribution characteristics of the receiver surface and the corresponding solutions are summarized. The non-uniform flux distribution on the surface of the receiver leads to a high local temperature, which has an adverse effect on the safe and efficient operation of the CSP system. Therefore, it is necessary to optimize and improve the concentrating collector system. For the four CSP methods, the main solution is to improve the concentrating performance and heat absorption performance to make the light-collecting match heat absorption.

Lidar is vital for the high-resolution and multi-parameter detection of concealed objects, objects in the ares like atmosphere, oceans, lands, and so on. Compared with the traditional time-domain or frequency-domain methods, time-frequency analysis can provide more insight into the analysis, interpretation, and processing of lidar signals. Time-frequency analysis has been widely used, including in feature analysis and extraction of atmospheric parameters, signal denoising, moving target imaging and detection, and micro-Doppler classification analysis. The methods used for the time-frequency analysis of lidar are further developed based on the basic principles and characteristics of time-frequency analysis.

Compared with master oscillator power amplifiers, fiber laser oscillators have the advantages of compact structure, low cost, strong anti-reflection light return capability, and good stability. With the development of fiber optic devices and processes, all-fiber laser oscillators achieve a near-diffraction limit output of 5 kW. As for the ytterbium-doped fiber laser oscillator, the research progress and problems faced by the space-coupled fiber oscillator and the all-fiber oscillator are introduced in detail. According to the nonlinear effects and mode instabilities that are the main limiting factors of high-power fiber laser oscillators, the technical approach to further increase the power of high-power fiber laser oscillators is discussed preliminarily, from the aspects of specially designed gain optical fibers and global oscillator optimization, in order to provide reference for the realization of single-mode fiber laser oscillators with 10 kW output.

The classification of medical images is a research hotspot in the field of computer-aided diagnosis and pattern recognition. Accurately categorizing human anatomical structure and lesion areas can maximally assist doctors in diagnosing diseases more accurately and quickly. Herein, for the particularity of medical images, typical medical image classification is first summarized and then described according to four aspects: image preprocessing, image segmentation, feature extraction, and classification. Next, the application of deep learning theory to medical image classification is introduced and discussed. Finally, the shortage of the existing medical image classification methods is addressed, and the development trend of the latest theories of deep learning in the field of medical image classification is discussed.

Modulation technology is a key technology to improve the performance of visible light communication (VLC) system. The fragility of high-speed LED visible light communication links and inter-symbol interference caused by multipath effect seriously affect the VLC system performance. To solve this problem, the various optical orthogonal frequency division multiplexing (O-OFDM) modulation technologies are proposed. An overview of O-OFDM modulation technologies is summarized and the proposed modulation technologies are classified into four types according to their modulation strategies. Taking different modulation technologies as examples, these strategies are used to analyze the mechanism and the VLC communication performance. In the end, the next step of work is prospected.

The silicon photonic arrayed waveguide grating (AWG) is an important device for achieving silicon-based photonic integration. We describe the research progress in the AWG based on different device structures and various materials. The device structure mainly includes conventional symmetric AWG, reflective AWG, cascaded AWG, and AWG with a multi-mode interferometer (MMI) aperture. Compared with the conventional symmetric AWG, the reflective AWGs have smaller footprint; the cascaded AWGs have better channel crosstalk performance; the MMI-AWGs can obtain a flattened spectrum response. The silicon nanowire AWG has a very small bending radius and makes devices more compact because of the high index-contrast of silicon; the silicon nitride AWG has good channel crosstalk and polarization performance. At the end of this paper, we describe the temperature and polarization insensitive AWGs, and forecast the future development trend of silicon photonic AWGs.

Micro-nanofiber Bragg grating (MNFBG) is highly sensitive and reliable to the changes of refractive index and concentration of the surrounding media due to its strong evanescent field propagation and wavelength-selective optical properties. The fabrication methods of MNFBG are illustrated, and the sensing principle of refractive index and concentration is analyzed. In addition, a research progress of refractive index and concentration sensors based on MNFBG is reviewed. Moreover, the methods for improving the sensing sensitivity of refractive index and concentration based on MNFBG are summarized, the existing problems in the current research are analyzed, and the development direction of refractive index and concentration sensors based on MNFBG is prospected.

Laser ranging plays an important role in the modern distance measurements, especially in the precise absolute distance measurement. In the applications such as the advanced manufacturing and space mission, an efficient distance measurement is required with a measurement range of tens of meters or up to a thousand meters but with a submillimeter or even micrometer precision. Simultaneously, the fast and efficient measurement is needed. It is a huge challenge to laser source for its flexibility, stability and traceability. A femtosecond optical frequency comb becomes a powerful tool in laser precision ranging due to its abundant modes, its relative stability of 10-11 and its traceability to a microwave clock. The high-precision time resolution of a femtosecond pulse in time-domain and the abundant optical modes in frequency-domain make the new measuring principles possible. The study status of femtosecond-frequency-comb-based absolute distance measurements is reviewed. The absolute distance measurement methods are classified according to the time-frequency characteristics of an optical frequency comb, and the principles and features of all kinds of methods are analyzed. Nowdays, the femtosecond-frequency-comb-based ranging possesses a micrometer or even sub-micrometer precision within a large distance range and also has a potential to simultaneously range multiple points. With the increase of measurement efficiency, the femtosecond-optical-frequency-comb-based ranging must and should be widely used in the large-size precision ranging.

The veiling glare index is a key criterion used to evaluate the image quality of optical systems. In general, the stray light coefficient can only be obtained with the actual measurement method after the optical system is processed. If results of actual measurement are unsatisfactory, then the method of suppressing stray light in the optical system must be adjusted to optimize its suppression effect. To overcome these questions, this study proposes a method for solving the veiling glare index during optical design. This method employs TracePro software to establish and analyze the optical system as well as trace rays. Matlab software is used to fit the simulation results and solve the veiling glare index. The proposed method can successfully calculate and analyze the veiling glare index in advance and accurately evaluate the stray light suppression of the optical system.

The identification of seed aging has a great significance for agricultural production and food security. In this paper, three kinds of legume seeds with different storage years are distinguished by a tri-step IR spectra method, including Fourier transform infrared (FT-IR) spectroscopy, second derivative infrared (SD-IR) spectroscopy and two-dimensional correlation infrared (2D-IR) spectroscopy. The results show that the FT-IR spectra of three kinds of legume seeds are very similar, which are composed of the absorption peaks of protein and carbohydrate. But slight differences of original infrared spectrum of legume seeds with different storage years are observed in the range of 1800-700 cm-1. The absorption intensity ratios of three kinds of legume seeds show a decrease tendency with the increase of storage time. The variance analysis of the absorption intensity ratio show that the absorbance ratio of the seeds with different storage years is significantly different. More obvious differences are exhibited in their SD-IR spectra in the range of 1800-700 cm-1. In 2D-IR spectra, the positions, numbers and intensities of auto-peaks and cross-peaks are obvious different in the range of 860-1690 cm-1 for broad bean seeds and red kidney bean seeds, and in the range of 1350-1800 cm-1 for soybean seeds. The numbers decrease with the increase of storage years and the intensities weaken with the increase of storage years. Partial least squares analysis can accurately distinguish legume seeds with different storage years. Three kinds of legume seeds show obvious spectra differences, which demonstrate that FT-IR spectrum combined with SD-IR spectrum and 2D-IR spectrum can be used to discriminate legume seeds with different storage years. The method can provide a simple and rapid spectral method for detecting seed aging.

The heavy metal in atmospheric particulates has great harm to human health, and the rapid detection of heavy metal elements therein is important. The laser-induced breakdown spectroscopy (LIBS) is applied to the rapid elemental analysis of atmospheric particulates, and the results show that atmospheric particulates contain Na, Al, Si, Cu, Mg, Fe and other elements. Taking Ca element as the reference element, we use the internal standard method to quantitatively analyze Pb element in the atmospheric particles. The calibration curve is obtained by fitting, and it is calculated that the limit of detection of Pb element is 34.3×10-6. Besides, the plasma temperature and electron number density of Pb element are calculated. The experimental results verify the feasibility of LIBS for the qualitative and quantitative analysis of heavy metal elements in atmospheric particulates and provide experimental evidence for monitoring atmospheric heavy metal pollution.