The performance of a multi-aperture coherent optical receiver based on the equal gain combining algorithm is analyzed for the satellite-to-ground downlink. Atmospheric turbulence, weather conditions, and aperture averaging effect are considered. The analytical model of bit error rate is deduced. Then the effects of the atmospheric turbulence, weather conditions, and aperture size on the sensitivity of the receiver are analyzed, which provides effective guidance for practical applications.

The network function virtualization (NFV) technology can significantly reduce the capital expenditure and improve the flexibility of network serves. The virtual network function (VNF) deployment problem is one of the key problems that should be addressed in NFV. A global constrained optimization model is established to minimize the maximum index of the used frequency slots, the number of used frequency slots, and the number of initialized VNFs,thereby solving the issues of routing and VNF deployment for service chains in inter-datacenter elastic optical networks. VNFs are divided into VNFs with and without dependency,respectively. To ensure the effectiveness of the proposed model, an efficient genetic algorithm based on highly efficient encoding, crossover operators, and mutation operators is designed. Several simulation experiments are conducted using different network topographies with different parameters. The experimental results denote that the proposed algorithm can obtain improved routing and deployment schemes for service chains in the VNF when compared with other algorithms.

A symmetry recovery-based receiver in an asymmetrically clipped optical orthogonal frequency division multiplexing (ACO-OFDM) system is proposed to achieve LED nonlinearity in visible light communication systems. Based on the symmetric recovery characteristics of ACO-OFDM, a spectral diversity that combines odd and even sub-channels can be achieved in the time domain. Theoretical and experimental results show that the proposed scheme has a good performance when it is applied to a highly nonlinear visible light communication system. Simulation results demonstrate that the proposed method outperforms conventional ACO-OFDM solutions. Furthermore, we find that the proposed method exhibits better performance when an increased modulation index leads to significant nonlinearity.

This study uses an end-to-end method for image deblurring based on deep learning to encode the blurred image and to subsequently decode it into a high-definition image. However, the lack of extracted feature information during encoding decreases the quality of the reconstructed deblurred image. To solve this problem, we propose two methods for improving the network structure. First, a dense network structure is added to the autoencoder network for extracting considerable feature information. Second, a multiscale perceptual field structure is introduced to extract considerable contextual feature information, comprising 4 scales of average pooling layers and up-sampling layers. The two improved methods achieve good image deblurring effects using the GOPRO and Kohler datasets.

The existing algorithms for detecting straight line segments are all designed for grayscale images. When implemented on color images, they discard the input image's color information; this is undesirable for accurately detecting straight line segments. To solve this problem, this study proposes an algorithm that directly detects straight line segments in color images. First, DiZenzo operator is utilized to compute gradient vectors in three color channels, and image edges are extracted based on the average gradient magnitudes and orientations. Then, the pixels on each image edge are tracked and they are linked or split to generate candidate straight line segments. Finally, based on the gradients of the points on the candidate line segments, Helmholtz criterion is used to eliminate the false line segments caused by noise, yielding the final set of line segments. Experimental results show that the new algorithm can fully exploit the color information of the input image, leading to significantly improved detection efficiency when compared with the existing algorithms. On the YorkUrbanDB image database, the highest accuracy of the proposed algorithm increases from 0.2207 to 0.2687 and a high F-score is achieved.

This study proposes a method which combines sliding filter with automatic region growth to extract surface crack defects on ceramic tiles with complex background interference and stereochemical structure. First, the acquired image is preprocessed and divided into two regions, namely tile head and texture area. Crack defects in the tile head region are detected by the customized sliding filter method, whereas those in the texture area are detected by the automatic region growth method. Finally, a morphological operation removes the spurious interference points and extracts the characteristic parameters of the cracks. Experimental results demonstrate that the proposed method effectively removes the complex background interference and extracts the surface cracks of ceramic tiles with stereochemical structure.

PointNet is one of the representative research results obtained from three-dimensional point cloud classification, which innovatively employs a deep learning model for point cloud classification and achieves good results. However, PointNet does not capture local information of each point, and it considers only the global features of point clouds. Herein, we propose a model for point cloud classification based on graph convolutional networks to solve this problem, in which a k-nearest neighbor (kNN) graph layer is designed and plugged into a PointNet model. The local information of point clouds can be effectively obtained by constructing the kNN graph layer in the point cloud space, which can improve the accuracy of point cloud classification. The point cloud classification experiment is conducted on the ModelNet40 dataset, and the effects of the different neighbor values of k on the output accuracy are compared. The results demonstrate that the highest classification accuracy is achieved when k is 20, reaching 93.2%, which is 4.0% higher than that of PointNet.

To solve the current issues of low effective matching rate and large stitching error of the remote sensing image mosaic algorithm, the matching and fusion/stitching processes of a remote sensing image mosaic algorithm are investigated herein. This algorithm utilizes a guided filter to preprocess the images. Based on the fast and efficient speed-up robust feature approach and the low-error bidirectional mutual selection matching strategy, the matching feature points are further purified by the random sample consensus algorithm; then, a homography transformation matrix is adopted to calculate the relative positional relationship between the images. In the fusion/stitching phase, the image fusion algorithm is modified according to Weber's law and the plant growth function. Finally, image fusion stitching is completed using the nonlinear S-type nonlinear fusion strategy. The simulations reveal improvements in the matching accuracy, average gradient of the image fusion/stitching resultant graph, and information entropy by approximately 1.01%-3.42%, 84.86%-146.26%, and 0.77%-2.22%, respectively. Thus, with respect to realizing high-efficiency and low-error remote sensing image matching, improved splicing quality and efficiency as well as high robustness of the fusion algorithm are achieved.

This study proposes an algorithm for detecting metal damage in combination with information entropy and low-rank tensor analysis to address the problems of low automation degree and recognition accuracy in the research of damage identification of metal parts. First, the image is denoised using the difference method, median filtering, and Fourier filtering. Second, according to the obvious difference between the damage of the metal part and its surroundings, the information entropy edge detection is used to obtain the edge information. At last, the low-rank tensor method is used to analyze the difference entropy and the weight entropy matrix to extract damage, and it is compared with other algorithms. The experimental results show that the algorithm can effectively and quickly identify metal damage with few noise points. The effective accuracy of the algorithm is higher than 80% with good robustness, which is higher than that of traditional algorithms.

Based on the tensor truncated nuclear norm and the spatial-spectral total variation regularization,a new model is proposed to realize super-resolution reconstruction of hyperspectral images to solve the problem that most hyperspectral images suffer from degradation in the acquisition process. First, two types of priori information in the hyperspectral images, i. e., the low rank-based priori information and sparse priori in the spatial and spectral domain, are explored. Next, using the low rank-based priori information in the spatial and spectral domain, a low-rank constraint model based on the tensor truncated nuclear norm is proposed to achieve a more accurate approximation of the rank function. Subsequently, using sparse priori information in the spatial and spatial domain, a spatial and spectral total variation regularization model is proposed to retain the sharp edges and more detailed information of the original image. Finally, the low-rank constraint model based on the tensor truncated nuclear norm and spatial and spectral total variation models are integrated. This new restoration model possesses the advantages of both the aforementioned models. The peak signal-to-noise ratio of 0.8 dB is obtained, and structural similarity indices are adopted to provide quantitative assessments of experimental results. The experimental results demonstrate that the proposed model achieves better visual quality than those of several existing related methods. The proposed model can effectively achieve the super-resolution reconstruction of hyperspectral images after being blurred and downsampled.

Region of interest is sliced smooth or polynomial, then accurate internal reconstruction can be performed by total variation (TV) minimization. The solution of TV minimization usually adopts the gradient descent method, taking the negative gradient of the objective function as the search direction, and then optimizes iteratively the objective function. In order to improve the efficiency of TV minimizing, this paper proposes a method to find the optimal solution direction by combining beetle antennae search (BAS) and gradient descent. The method selects the gradient descent direction or the optimal solution direction which is based on the individual “left and right whiskers” to iterate, according to the generated random number and threshold during the TV minimization process. The simulation experiment and the actual experiment show that the proposed algorithm has fast convergence speed and good reconstruction effect.

To achieve good stereoscopic effect and avoid visual discomfort, an adaptive horizontal parallax adjusting method is presented for stereoscopic images. The depth information in a scene is extracted by feature point matching, obtaining the maximum and minimum depths. An appropriate baseline is obtained after the resetting of baseline based on the current baseline and the maximum and minimum parallaxes acceptable to the human eye, ensuring that the scences are imaged in a comfortable stereo imaging area. The proposed method does not introduce stereo image distortion. The control effect of the proposed method is verified by experiments. Experimental results show that the proposed method can adaptively adjust the parallax and realize strong stereoscopic effect.

A constrained link routing algorithm is proposed to solve the problem of unconfigured dilated Benes optical switching chips. First, the algorithm groups the optical switches to be configured according to their lateral constraint and subnet connection relationship, and then it optimizes the optical switch combination state to form a link route that satisfies the switching requirement. In this manner, the configuration efficiency of optical switching chips is improved since no processing is required for other idle optical switches. Taking a 16×16 dilated Benes optical switching network as an example, the execution process of the constrained link routing algorithm is described and compared with other configuration routing algorithms. Furthermore, the role of the algorithm in improving the fault tolerance of a chip when its optical switch loses the switching or light-passing function in the chip network is analyzed.

In allusion to the difficult-to-remove oxide layers on a titanium rolled plate surface, a process for laser cleaning the oxide layer on a titanium rolled plate surface is investigated using a laser cleaning machine with an average power of 200 W. Based on the law of the temperature rise of the metal oxide under the action of the laser, raising the laser single-pulse power density is proposed as an effective solution to the problem of laser removal of the oxide layers. Combined with the results of energy-dispersive X-ray spectroscopic analysis and the cleaning effect of the oxides on the surface of titanium rolled plate under different laser cleaning processes, the oxide on the surface of the titanium rolled plate is shown to be effectively removed using laser cleaning when the laser power density reaches 4×10 8 W/cm 2. Noticeably, this process does not affect the tensile properties of the material.

Experiments on the laser-surface remelting of vermicular-graphite cast-iron valve seats are performed in this study using a Nd∶YAG solid-state laser with a rated power of 3 kW. The macroscopic morphology, microstructure, hardness distribution, and cracks of the remelting layer of the laser-surface remelting of the vermicular-graphite cast-iron valve seats are analyzed using a scanning electron microscope, microhardness tester, and phosphor flaw detection method. The results show that after the surface of the vermicular-graphite cast-iron valve seats is remelted using a laser, the dendritic structures appear in the central region of the remelting layer and the needles and granules are dispersed among the dendritic structures. The morphology of the interface between the remelting layer and substrate is serrated, and the bivalve structure of the martensite and ledeburite shells is observed in the heat-affected zone because of factors such as the diffusion of graphite reducing the melting point of the material in the vicinity of the graphite and rapid heating and cooling of the laser. The hardness from the remelting-zone surface to the substrate gradually decreases because of the differences in the heating and cooling rates in the different regions of laser remelting of the valve seats. The cracks in the laser remelting layer of the valve seats can be restrained by adjusting the laser parameters. A remelting layer without cracks on the surface of the vermicular-graphite cast-iron valve seats of laser remelting can be obtained under the laser remelting process with parameters of laser power P=500 W, defocusing amount Δf=-1 mm, and laser scanning velocity V≤5 mm/s.

The electrochemical corrosion behavior of laser-cleaned AH32 steel in 3.5% (mass fraction) NaCl solution was studied herein through the electrochemical method to investigate the effect of laser cleaning (LC) on the corrosion resistance of steel. The polarization curve, electrochemical impedance spectroscopy, and surface morphology of the sample were tested and analyzed. Results show that the corrosion resistance of AH32 steel is improved post LC. Within the material damage threshold, the self-corrosion potential of AH32 steel moves in a positive direction with increasing laser energy density and spot overlap rate; simultaneously, the corrosion current density decreases, and the impedance arc becomes larger, thereby the corrosion resistance increases. The corrosion resistance of AH32 steel is the best when the laser energy density is 20 J/cm 2, and the spot overlap ratio is 50%. These results can provide a reference for studying surface properties post nanosecond pulse LC.

We build a chaos synchronization communication system with less environmental impact and better stability, using the dual-path mutual coupling semiconductor laser as a transmitter of chaotic light. The communication capability of this system is studied from the hiding of time delay characteristics, the correlation between the transmitting and receiving lasers, and the output chaotic optical bandwidth. The peaks of the feedback time delay characteristics of the external cavity of this system, the peaks of the mutual information function of the transmitting and receiving lasers, and the output chaotic optical bandwidth of the system are calculated by using mathematical tools such as autocorrelation function and mutual information function. The peak value of feedback time delay characteristic of the dual-path mutual coupling system can be reduced to 0.1899 when the transmitting and receiving lasers are synchronized by adjusting the parameters, which is 0.1696 lower than the characteristic peak value of a single-path mutual coupling system under the same conditions. The maximum cross correlation coefficient between receiving and transmitting lasers is 0.9113. In addition, the central frequency detuning up to 21 GHz can broaden the output optical bandwidth to 3.319 GHz. The experiment verifies that the chaotic carrier generated by the dual-path mutual coupling system can hide the sinusoidal signal well and successfully demodulate at the receiver.

This study proposes a face detection algorithm based on a convolutional neural network considering the scenario of a classroom, where the faces of students sitting in the back rows might not be visible. First, the algorithm extracts face features in two stages using a residual neural network. Then, it builds a feature pyramid and combines the Softmax loss function with center loss function to train a face recognition model based on a proper activation function. Upon applying the algorithm to the Wider Face dataset, it achieves an accuracy of 95.2% and mean average precision values of 93.0%, 87.3%, and 58.3% for three levels of validation sets, respectively.

This study proposes a video smoke detection method combining the Gaussian mixture model (GMM) with a convolutional neural network (CNN) to ensure real-time and accurate video smoke detection in complex scenarios. First, background subtraction based on GMM and morphological methods are used to extract motion objects from video images. Second, a CNN model for video smoke detection is designed, taking into account the limitations of smoke detection efficiency and overfitting of the network model. Finally, the designed CNN model is trained and tested by using positive and negative smoke sample images. The output probability threshold of the network model of motion objects is set reasonably, which can effectively remove the non-smoke interference items that are not covered in the training samples. The false alarm rate can thereby be reduced. Experimental results prove the validity and feasibility of the method. The accuracy of video smoke detection reaches 97.5%, and the average response time of the smoke alarm is 4.58 s, satisfying the real-time demand of smoke detection in complex scenarios.

In this study, we propose an algorithm based on machine vision to detect heel-strike events for solving the problem that the gait recognition technology based on wearable sensors is considerably dependent on the cooperation of participants, with high energy consumption and harsh application conditions. The proposed algorithm can accurately detect heel-strike events using ordinary cameras without the cooperation of participants. Initially, we develop an innovative feature for representing gait patterns by designing a consecutive-silhouette difference map (CSD-map). A CSD-map can encode the binary image of several consecutive pedestrian contours extracted from the video frames and output the combination as a single feature map. Different numbers of consecutive pedestrian contour differences result in different types of CSD-map. Further, a convolutional neural network is used for feature extraction and classification of the imaged heel-strike events. In a public database of video data for training and testing, we find 124 individuals under five angles in different wear conditions, and the experimental results obtained using these images denote the accuracy of our method. The identification accuracy is observed to be greater than 93%.

To solve the problem of mismatch of the edge region in the local stereo matching algorithm, a cross-scale local stereo matching algorithm based on edge weighting is proposed. In the cost computation stage, an edge similarity measurement method is proposed according to the number and structural information of edge points, and the points satisfying the constraint conditions are weighted by two strategies. In this way, the recognition of corresponding points in the target and reference maps are improved. Cross-scale model is introduced in the cost aggregation stage, and guided filtering is used for aggregation. Finally, the disparity map is obtained by disparity computation and refinement. Four sets of standard stereo image pairs and 27 sets of extended stereo image pairs are tested on the Middlebury benchmark. The average mismatch rate of non-occlusion regions is 7.88% without any refinement steps. Experimental results show that the proposed algorithm effectively improves the matching accuracy of the edge region.

An improved algorithm based on feature pyramid networks is proposed for small target detection. A prediction optimization module is introduced, which is combined with the context information of the region of interest to make the feature information more robust, multi-threshold prediction networks with internal cascade are predicted, and the multi-scale and multi-stage prediction is realized finally. On the premise that the network parameters are basically unchanged, the accuracy is further improved. The experimental results show that the accuracy of the proposed algorithm reaches 80.9% in the VOC2007 test after the training of the standard data set VOC07+12, which has good detection performance.

Camera calibration is a very important part in machine-vision-system research. To calibrate a camera better and faster, we propose a camera-parameter-optimization method based on hybrid particle-swarm optimization. First, we obtain the internal and external parameters of the camera by the least-squares method and use them as the initial values of the parameters to be optimized. Then, we establish the objective function using the minimum-distance criterion. Next, we use a hybrid particle-swarm optimization algorithm to further optimize the camera parameters, and finally we obtain the camera parameters with only small errors. Our experimental results show that the optimization algorithm can converge quickly and accurately. Therefore, this method is able to improve camera-calibration accuracy of the camera to some extent.

Robot movement cannot be accurately estimated because of the impact of moving objects in a dynamic environment. Therefore, this study proposes a visual odometry algorithm based on ORB (Oriented FAST and Rotated BRIEF) feature regional segmentation. Further, using the distance invariance of the feature points in the adjacent regions in a three-dimensional space, the extracted feature points are segmented and the feature points generated by the moving objects in the image are separated from the feature points in the static background, and influences of dynamic object feature points are removed. Subsequently, the position of the camera can be estimated, thereby removing the interference caused by the dynamic objects in a scene. The experimental results show that the visual odometry algorithm based on ORB feature regional segmentation can perform real-time pose estimation in both dynamic and static environments with good robustness and high precision.

The demand for high-power light-emitting diode (LED) lighting shows an increasing trend; however, with the increasing power of blue LED chips, problems such as heat dissipation and light attenuation are increasing as well. To explore high-power LED special lighting, an array-based blue LED is used to excite Ce∶YAG fluorescent crystal using an innovative structural device to output high-lumen-density light on small-area end-faces. Many factors affect LED light emission, such as the absorption of blue light by crystal rods with different doping concentrations, the differing light-emitting effects of high-powered LED excitation for crystal rods with different Ce-ion-doping concentrations, and the influence of end-face treatment of crystal rods upon the light output. The result shows that a substrate made by aluminum nitride satisfies the heat-dissipation requirement of a high-powered LED; the use of crystal rod with a doping concentration of 0.8% and the end-face treatment of the crystal rod can greatly increase luminous flux. The experimental device can output more than 6000 lm of light flux with a lumen density up to 1640 lm/mm 2, which can be applied to medium- and high-brightness lighting equipments.

Based on the optical analysis software of LightTools, the transmitting processing of light in CD-pattern microstructure diffuser is analyzed. The injection molding process is used to verify the effect of this model. The effect of CD pattern microstructure on the light extraction efficiency is studied. Through theoretical research, simulation, and experimental comparison, the surface microstructure of the diffuser is modeled and the ray tracing is carried out on a large scale. The results show that, in the 80 mm×80 mm illumination area, the optical lumen efficiency of the CD pattern microstructure diffuser is obviously improved compared with that of the non-microstructure diffuser, and the line width of 200 μm can increase the optical lumen efficiency by 13.52%. According to the optimal simulation size, a CD pattern microstructure diffuser composed of several concentric rings is prepared, and an 8-inch LED downlight is used as the test luminaire. The total lumen output of the micro-structure diffuser, general diffuser, and luminaire without the diffuser is measured by the integrating sphere. The results show that the CD-pattern micro- diffuser has a transmittance of 90.6%, and its light transmittance is 8.8% higher than that of the general diffuser.

Raman lasers generated by optical injection locking can be used in atom interferometers. However, phase noise significantly affects the phase of atom interference fringe; thus, we experimentally study the phase noise of the Raman laser under varying conditions. Two phase-locked Raman lasers exhibiting a frequency difference of 6.834 GHz are obtained by modulation of optical fiber electro-optic modulation (EOM) and injection locking. In the experimental system, the master laser is modulated using a 6.834-GHz fiber EOM, and its output is emitted as seed light. By adjusting the slave laser, the first amplification output beam of -1 level frequency-shifted laser can be obtained. The result demonstrates that the Raman laser beat linewidth is not wider than 3 Hz, the frequency is adjustable to 300 MHz, and the phase noise is less than -60 dBc/Hz from 10 Hz to 100 kHz when the master laser is frequency-locked.

A portable light source and power meter system has been designed based on a digital supermode distributed Bragg reflection (DS-DBR) laser and an InGaAs photodetector. The light source module of the system uses both the OPA569AIDWP chip and the ADN8834 chip as the controlling core, providing excellent control of the wavelength accuracy and power stability. The power meter module employs the multi-grade amplification range control and insulation denoising technologies along with the software algorithms, which significantly suppresses the system noise and improves the sensitivity and dynamic range of the power meter. The test results demonstrate that the light source module can effectively control the output light wavelength. Also, the side-mode suppression ratio of the output light is higher than 50 dB, and the linewidth is smaller than 1 MHz with satisfactory output quality. Furthermore, the dynamic range of the power meter module reaches 80 dB, and the linearity is increased to 0.08 dB. In addition, our system adopts an embedded control technology to allow human-computer interaction; thus, it can be applied to various settings such as laboratories and industries. Therefore, with such attractive and broad market prospects, the application range of our portable system is considerably broadened.

In this study, we propose a scheme to transform quantum entangled states. In this scheme, a quantum controlled-not (CNOT) gate is realized in coupling systems of the nitrogen-vacancy (NV) color center in diamond and microtoroidal resonators (MTR). A quantum entangled W-state is converted to an entangled GHZ-state based on the cross-Kerr nonlinearity. The scheme displays a high conversion efficiency under the current experimental conditions. This may effectively assist in the preparation of quantum entangled states and the implementation of quantum information processing.

The transmission distance of a self-referenced continuous variable quantum key distribution (SR-CV-QKD) protocol still needs to be improved. Therefore, this study proposes an optical amplifier to improve the system and effectively compensate for the attenuation of phase noise caused by reference pulse. The secret key rate is derived through security analysis. Simulation results based on practical parameters show that adding an amplifier can improve the performance of the SR-CV-QKD protocol with a weak reference pulse and extend the maximum transmission distance.

A quantum key distribution based optical network is a network formed by the interconnection of quantum key distribution devices through optical fibers. In the quantum key distribution optical networks, the key pool combines the characteristics of gradual accumulation and instantaneous consumption of secret keys to realize the efficient storage and management of key resources. This study analyzes the characteristics of quantum key distribution technology and the functional significance of the key pool. In addition, this study proposes a design of the “key as a service” framework and the quantum key distribution optical network architecture based on the key pool. A key pool construction method and a key resource scheduling method based on the architecture are proposed, and they are verified by a network simulation.

The research on three-dimensional (3D) reconstruction after the processing of point-cloud data from airborne lidar has received considerable attention for a long time. Because of the deficiencies of greedy triangulation (GT) algorithm for the 3D reconstruction of stockpile point-cloud data, a GT algorithm based on inverse-distance-weighted interpolation is proposed herein. This proposed algorithm can obtain data of unmeasured area; further, it can acquire uniform data by controlling the interpolation area. The proposed GT algorithm is used for the 3D reconstruction of stockpile in the spot, and comparison experiments show that the proposed algorithm produces results that are closer to reality and that it has strong application potential.

A multi-scale saliency detection-based visual attention mechanism is introduced to eliminate noise and enhance the quality of the hyperspectral images. Further, a hyperspectral image classification method is proposed by combining the clustering dimensionality reduction and visual attention mechanism in accordance with the hierarchical clustering algorithm. Subsequently, dimensionality reduction, acquisition of saliency mapping, and support-vector-machine-supervised classification experiments are conducted by considering the Indian and Pavia datasets as examples. The results denote that the proposed method can considerably improve the classification accuracy and efficiency of hyperspectral images.

By combining the wavelength demodulation and second harmonics, the trace H2S gas is detected based on the quartz-enhanced photoacoustic spectroscopy (QEPAS). A distributed feedback semiconductor laser with an output wavelength of 1578 nm is used as the excitation source. Furthermore, a finite element analysis method is used in the simulation calculation for a quartz tuning fork, typically used in QEPAS systems, to obtain its first six modes and resonance frequencies. In an experiment, an acoustic micro-resonator with a length of 4 mm and an inner diameter of 0.7 mm is added, and a transimpedance amplifying circuit is optimized to detect the H2S gas under optimal experimental conditions. The results denote that the second harmonic signal of the QEPAS system exhibits a strongly linear relation with the H2S gas concentration, and a detection limit of 19.3×10 -6 is obtained.