In this study, a confocal system is proposed for which super-resolution imaging can be achieved based on the negative refraction and subwavelength imaging characteristics of two-dimensional photonic crystals (PCs). Further, the focusing and imaging processes of a confocal system are simulated with the finite-difference time-domain method. When the distance from the focus to the lower surface of a PC becomes 1.55 μm and when the transverse coordinate x=4, the full width at half maximum (FWHM) of the focus becomes 0.593λ, which is less than the incident wavelength; further, the FWHM of the reflected light in the right of the image point becomes 0.496λ, which indicates that super-resolution imaging can be realized. The FWHM of the image point decreases continuously when the focus shifts to right; further, the axial resolution of the confocal system becomes 2.2λ when the pinhole and focus are fixed.

For a fiber having an inverted parabolic distribution as its refractive index distribution, adding a low refractive index layer between the core and the fiber cladding can create an improved novel fiber. This novel fiber can accommodate a high-order orbital angular momentum (OAM) mode field. Calculation of the normalized cutoff frequency showed that the novel fiber can support 9 fiber modes and accommodate the OAM modes with a topological charge of 3. According to the calculations, the effective refractive index differences for the degenerate modes of LP11 and LP21 both reach 2.1×10-4, and the effective refractive index difference between the degenerate modes of LP31 reaches 10-3. The vector modes were effectively decoupled and the degenerative separation effect of the higher order modes was improved. Finally, the influence of the low refractive index layer's distribution on the optical fiber transmittable vortex light mode was analyzed.

Based on nonlinear Schr dinger equation, we research the light field properties, including the distribution and the stability, and dynamical behaviors of multi-polar lattice solitons, distributed on a single channel or across different channels in nonlinear media with optical lattices. These survey results indicate that there are two kinds of multi-polar solitons such as in-phase soliton and out-of-phase soliton in such systems. Significantly, multi-polar lattice soliton distributed on a single channel can only be out-of-phase, but an in-phase multi-polar lattice soliton is not allowed. Among three kinds of multi-polar lattice solitons, stability of in-phase multi-polar lattice soliton is the worst, and that of out-of-phase multi-polar lattice soliton distributed across different channels is the best. If optical lattice is removed, three kinds of multi-polar lattice solitons present different dynamical behaviors. After the optical lattice is removed, an in-phase multi-polar lattice soliton first splits into a corresponding number of light beams, and then they merge into one beam, or their energy is redistributed, or they collide periodically. Nevertheless, the out-of-phase multi-polar lattice soliton splits into several light beams, and then the beams repel each other, in which two side light beams transmit forwards along their respective straight trajectories, and the angle between any of their straight trajectories and its original direction of propagation is equal; especially, light beams, generated from the fission of out-of-phase multi-polar lattice soliton with a small propagation constant, may diffuse when the optical lattice is withdrawn. These results may have certain value to all-optical controlling techniques such as all-optical routing and all-optical switching.

In order to solve the difficult problem of low target recognition rate caused by image blur in the laser rang-gated imaging process, we propose the keep-feature convolutional neural network (KFCNN) model for the target recognition of laser rang-gated images. Different from the convolutional neural network (CNN), the KFCNN model is used to improve the recognition rate of blurred targets and the robustness of target recognition with a new keep-feature layer. To achieve keep-feature in the KFCNN model, we optimize the keep-feature objective functions and the training by imposing keep-feature constraints and regularization. In addition, the feature maps of training samples are kept consistent before and after image blur when the value of the keep-feature objective function is reduced. The experimental results show that KFCNN improves the problem of recognition rate reduction caused by image blur and further improves the recognition rate of specified targets in laser rang-gated imaging.

Based on the characteristics of multi-band, inter-band information redundancy and spatial information correlation of hyperspectral images, a spatially-correlated and semi-supervised feature extraction (SCSSFE) algorithm with locality preserving projection (LPP) is proposed. This algorithm defines a new pixel weight calculation function for the different spectral characteristics with the same objects and the different objects with the same spectral characteristics to preserve the spatial distance and the spectral similarity of hyperspectral image by means of the neighbor structure in image space and the intra-class and inter-class discriminant weights. Then, the features of hyperspectral images are extracted by the weight function combined with LPP. Thus the similarity among the same objects and the discrepancy among the different objects are maximized. The proposed LPP-SCSSFE algorithm is verified through the hyperspectral image classification experiments on the two datasets of Indian Pines and Pavia University. The highest overall classification accuracies of the LPP-SCSSFE algorithm reach 87.50% and 91.29% for the respective datasets, better than those of the existing feature extraction algorithms. These results indicate that the spatial correlation and the spectral similarity of hyperspectral images are fully taken into account in the proposed algorithm, and thus the more representative features are extracted and the classification accuracy is enhanced.

Aiming at the known deficiencies with complex training, strict parameter-tuning skills and experiences, difficult theoretical analysis of deep neural networks, an improved image classification algorithm with high training efficiency, strong interpretability and simple theoretical analysis is proposed, in which the principal component analysis network (PCANet) is used for feature extraction and the flat neural network (FNN) is for classification. In addition, the model parameters can be obtained by direct calculation and the flat neural network adaptively determines the number of nodes according to the training dataset. When the nodes increase, it is not necessary to retrain the model and only the parameters need to be adjusted locally to update the model. The experimental results show that the proposed model can acquire rapid training. Moreover, it possesses more competition in recognition accuracy compared with other unsupervised classification algorithms and traditional deep neural networks.

The Criminisi algorithm, as one representative of excellent image inpainting algorithms, can used to obtain a better visual effect when partially damaged images are inpainted, but when this algorithm is used to perform the matching block search, the matching range is too small because the amount of information provided by the blocks to be repaired is less during the matching block search. For this problem, an improved Criminisi image inpainting algorithm based on rough data-deduction is proposed, in which rough data-deduction can be used to expand the search space, increase the search data, expand the search scope, and deepen the search depth. The proposed algorithm has some improvements in the search rules. The image content is divided into a dataset according to the structural information of images. The amount of pending repairing information is extended by rough data-deduction. The matching block search range is expanded. Based on these, the matching blocks are searched and the broken images are repaired. The results show that compared with the traditional Criminisi algorithm, the improved algorithm can be used to expand the matching block data sizes, search more data, obtain better visual effects, and improve the peak signal-to-noise ratio of images.

In order to solve the problem that image quality is easily overfitted by a single model, a hyperspectral image quality evaluation algorithm is proposed based on multi-model fusion. Taking image noise, ambiguity and cloud content as the degraded features, a remote sensing image subjective evaluation database is established. The support vector regression method and the integrated decision tree method are respectively selected to establish a quality evaluation model for training set images with evaluation values. The image quality evaluation results based on model fusion are obtained via linear regression fitting of the two single model evaluation results. At the same time, the generalized regression neural network model is introduced as a reference, and several models are compared from four aspects of mean square error, regression fitting index, classification accuracy and training time. The experimental results show that the proposed model fusion algorithm has relatively high fitting accuracy, relatively strong generalization ability and relatively little training time.

Zero walking amount is one of the important indicators to evaluate the stability of sight. However, when the zero walking amount is measured, the influence of the repeated loading error of sight on the zero walking amount is usually neglected, because the ordinary sight has a low requirement of zero walking. In contrast, in the zero walking amount measurement of a high-precision sight, the repeated loading error cannot be ignored. Based on the principle of self-collimator, a device for automatically eliminating the repeated loading errors is designed, which consists of an automatic laminating device with a semi-reflecting and semi-transmitting and an electro-optical autocollimator. In addition, a pressure sensor is combined with a piezoelectric ceramic transducer (PZT) micro-displacement device to form a micro-motion fine-tuning closed-loop feedback system and the real-time adjustment of the pose of the semi-reflecting and semi-transmitting is realized. Thus the semi-reflecting and semi-transmitting and the end surface of the objective lens are completely fitted. It has been verified by experiments that the device can be used to achieve a measurement accuracy of 2" in the 5° measurement range, which is sufficient to meet the measurement requirements of zero-level reloading errors for a high-precision sight.

To measure the transmission wavefront of a large-aperture aspheric lens, we propose a method by using a computer-generated hologram (CGH) as a phase compensation element. The transmission wavefront of a large-aperture aspheric lens is typically tested by a sphere interferometer. To conduct a null test, phase compensation is required to compensate for high-order aberrations (HOAs). However, validating the accuracy of conventional refractive-phase compensation elements is difficult, and they therefore generate unreliable measurement results. An alternative type of phase compensation employs a CGH, which is a diffraction element that can also be used to conduct a null test. To conduct a null test for the transmission wavefront of a large-aperture aspheric lens, we design and manufacture a CGH. A contrast test is implemented with two different phase compensation elements, which results in two measured data those are very similar to one another, between which the peak valley (PV) difference is 0.034λ (λ is the wavelength of the detect light), and the root mean square (RMS) difference is 0.006λ. Therefore, the null-test method using a CGH proposed in this study displays a high degree of accuracy.

The accurate and fast extraction of structured light stripe centers is a key problem in a three-dimensional (3D) measurement system. Aiming at the existing contradiction between extraction precision and speed of structured light stripe centers, a novel laser stripe center extraction method is proposed based on the Hessian matrix and the regional growth. First, the adaptive threshold method is used to extract the region of interest from the images, and the initial position of the pixel-level stripe center is determined by the maximum value of the gray value. Second, the sub-pixel-level strip center point in the normal direction of the initial point is obtained by the Hessian matrix. Finally, the strip center is used as a seed point for the regional growth iteration operation and thus the stripe center is accurately extracted. In the regional growth algorithm, the problem of a large number of Gaussian convolutional operations in the traditional method is solved, and the extraction speed of the stripe center is increased. The experimental results show that the stripe center extracted by the proposed algorithm has a high accuracy and the real-time online requirements of the 3D measurement system is satisfied. The mean square error (RMS) of this algorithm is reduced by 2.02 pixel compared with that of gray-gravity algorithm, and the extraction speed is 40 times higher than that of Steger algorithm.

This study analyzes the effects of image noise on pose estimation error to meet the requirements of high-accuracy pose measurements in high-speed complex flow fields. A monocular pose estimation optimization algorithm of adaptive reprojection error is proposed based on the traditional nonlinear optimization pose algorithm. In the algorithm, the initial value of pose estimation is set as the center, wherein the constraint interval is set. A new penalty function is formulated to transform the constrained nonlinear optimization into an unconstrained nonlinear optimization. The relation between the image reprojection error and constraint interval is analyzed; then, the corresponding mathematical model is built based on the analysis results. The constraint interval is automatically adjusted according to the model to optimize the parameters of pose estimation for the constrained nonlinear adaptive reprojection error. The simulation results reveal that the proposed algorithm provides an optimal solution for the reprojection error and pose estimation parameters under different image noise levels. The proposed algorithm is superior to the traditional nonlinear optimization algorithm and has higher pose estimation accuracy.

As traditional fault identification methods typically exhibit considerable processing complexity, are often time-consuming, and display a low degree of precision, a novel approach based on wavelet packet analysis using a support vector machine (SVM) is proposed in this study for the automatic identification of fiber defects in optical time domain reflectometry (OTDR). OTDR is initially used to acquire the original data of the fiber under test (FUT). Further, the event signs are decomposed by the optimal basic wavelet packet after the events are located, and the normalized energy features of the event signs as eigenvectors are extracted as input of training and testing based on the results of signal reconstruction. Finally, the SVM model is built,and fiber defects can be identified with the eigenvector as input. Subsequently, the SVM identification technique is used to obtain effective classification of the events as either reflection events, which are caused by connectors, or as non-reflection events, which are caused by bent events. In this study, two classification tests have been performed on a total of 2500 reflection and non-reflection events in airborne optical cable samples. The experimental results indicate that our method achieves a recognition rate of 99% in 3.03 s when the number of training samples is 1750 and when the number of testing samples is 750. Additionally, the recognition rate is increased by 2% and the recognition time is observed to be only 1% when compared to the previously proposed recognition method that is based on the backpropagation neural network. At present, the proposed method is successfully applied in the field detection equipment of airborne optical cable components independently developed.

To improve the interface, shorten the process, and reduce copper consumption as well as production input, we use selective laser sintering technology to rapidly prepare molded parts of porous graphite. The effects of graphite preformation states, electroplating process parameters, and plating time on the deposition rate, the apparent morphology and bonding conditions of the plating and substrate are studied. The results reveal that for the preformed graphite that is not impregnated with phenolic resin, the bonding strength between the plating and the substrate does not meet the application requirements. After the preformed graphite is immersed in phenolic resin for 2-3 times under a vacuum pressure, the plating deposition rate is best, and the bonding force between the plating and the substrate meets the application requirements. When the electroplating time is 30 min to 50 min, the deposition rate is best, and a complete, continuous, and glossy plating on the surface can be obtained. When the current density is 5×102-6.25×102 A·m-2, the plated grains are uniform and dense, and a better plating can be obtained.

Owing to its rotational symmetry and imaging invariance, a sphere has unique advantages when used as a spherical target in binocular vision systems. However, the image of a sphere on an image surface is generally an ellipse, rather than a standard circle, and the center of the ellipse does not coincide with the center of the sphere, resulting in errors when the location of the sphere's center is determined. To calculate accurate image coordinates of the center of a sphere to improve positioning accuracy, imaging models of spheres in space are established and analyzed, and a high-precision positioning method for the image of a sphere's center is presented based on the pinhole imaging model and perspective imaging principle. Combined with the focal length of the camera, the image coordinates of a sphere's center can be calculated after distortion correction of the edge points. The influence of edge noise and incomplete edge on this method is simulated and analyzed herein. Finally, the experimental results show that the proposed method can significantly improve the positioning accuracy of an image of a sphere's center; therefore, the proposed method is suitable for the positioning of a spherical target in a binocular vision system and for other applications requiring the accurate positioning of the center of a sphere.

Focusing on the target tracking drift and loss problems under severe occlusion for the traditional tracking methods, a long-term robust target tracking algorithm is proposed in the framework of Kernelized Correlation Filter (KCF) tracking. A combined confidence measurement method including occlusion information is introduced during the tracking process and used for the robust updates. If the result of the confidence graph by the KCF algorithm indicates that the target is occluded, a block mean shift (MS) algorithm is introduced to track this target and the local information is used to obtain the final location of this target. The performance of this algorithm is tested based on the eight sets of video sequences in the OTB-13 test library. The accuracy is increased by 0.7% and the success rate is increased by 5.7% compared with those of the traditional KCF algorithms. The test results show that even when the target is seriously occluded, the proposed algorithm still has a good tracking effect and a long-term stable target tracking is realized.

Aiming at the problem that the existing person re-identification algorithm cannot be adapted well to the variances of illumination, attitude and occlusion, a novel person re-identification algorithm based on feature fusion and subspace learning is proposed, in which the Histogram of Oriented Gradient (HOG) feature and the Hue-Saturation-Value (HSV) histogram feature are first extracted from the entire pedestrian image as the overall feature and then the Color Naming (CN) feature and the two-scale Scale Invariant Local Ternary Pattern (SILTP) feature are extracted in a sliding window. In addition, in order to make this algorithm have better scale invariance, the original images are first down-sampled twice and then the above features are extracted from the sampled images. After the features are extracted, a kernel function is used to transform the original feature space into a nonlinear space, in which a subspace is learned. Simultaneously, in this subspace, a similarity function is learned. The experiments on three public datasets are conducted and the results show that the proposed algorithm can be used to improve the re-identification rate relatively well.

Based on the structural characteristics of laser deposition formed metal thin-wall workpieces, the confirmatory test of subarea annular scan is conducted to reduce the temperature gradient of laser deposition formed layers and the residual stress of workpieces. The experimental results show that the usage of subarea annular scan paths can reduce the residual stress of formed parts effectively. An algorithm for generating the subarea annular scan paths is proposed. First, in order to avoid warping deformation caused by the long line scan, a subarea algorithm based on the extreme point search within the scan line limit length threshold is proposed according to the scan line limit length of laser deposition forming, and used for partitioning the independent contour group. Then based on the structural characteristics of thickness-changeable thin-walled workpieces and according to the optimization criterion of wall-thickness, the midline scan and annular scan are combined as an algorithm for the generation of subarea adaptable annular scan paths. Meanwhile, in order to avoid the over-cladding of the transition zone between scan rings, the endpoint offset algorithm is adopted to achieve a smooth transition connection between scan rings. Finally, in order to reduce the temperature gradient of layer surface and the residual stress of workpieces, a sequential optimization algorithm for subarea scan is proposed based on the four-color theorem.

The in-situ WC-reinforced nickel-based alloy ceramic coating is prepared on the surface of a cold die steel with a fiber laser. The inverted metallographic microscope is used to observe the macroscopic morphology of cracks in the cladding layer and the variation of the ice crystal WC with specific powder, specific energy and the powder feeding voltage. In addition, a microhardness tester is used to measure the microhardness of cladding layers. The results show that when the voltage ratio between WC and Ni60 powder exceeds 3∶2, the cracks in the cladding layer cannot be eliminated by the adjustments of specific powder and specific energy. The microhardness of the cladding layer increases with the increase of specific energy, while decreases with the increase of specific powder. Moreover, the increase of the cladding layer conforms to a polynomial distribution accuracy of up to 96%. In contrast, when the powder feeding voltage ratio between WC and Ni60 is lower than 3∶7, WC has nearly no effect on the microhardness of the cladding layer. However, when the powder feeding voltage ratio exceeds 4∶6, the microhardness jumps to 780 HV. According to the influence of the ratio between specific powder and specific energy on the increase and melting depth of the substrate, the process parameters of laser cladding are adjusted to eliminate the cracks of the cladding layer and simultaneously maintain the microhardness at about 800 HV.

The kinetics and defect trapping at the solid-liquid interface of simple metal Ag during the solidification process are investigated by molecular dynamics simulation. It is found that there exists a certain characteristic value (T*) for the interfacial temperature of metal Ag at which the growth rate reaches a maximum value. Meanwhile, the vacancy defects are predominant in the solidification process of liquids. The calculation results show that there exists a linear relationship between defect concentration and interfacial temperature. The defect concentration gradually increases with the decrease of interfacial temperature, and a transition occurs near this characteristic temperature. In addition, the defect concentration is found to be dependent on growth velocity. Above the characteristic temperature (T>T*), the defect concentration almost linearly depends on the growth velocity, and both are independent on orientations. In contrast, below the characteristic temperature (T<T*), an obvious anisotropy occurs along the (100) and (110) orientations, indicating that the defect trapping is associated with the growth mechanism transition.

To ameliorate the serious sidelobe noise and high time complexity of the conventional synthetic-aperture algorithm in medical ultrasonic endoscopy systems, a coherence-factor-based real-time synthetic-aperture algorithm is proposed in this paper, which is realized with the high-speed parallel processing capabilities of a graphic processor (GPU). Introducing the coherence factor effectively suppresses sidelobe noise in an ultrasonic echo image, and the algorithm is processed using a GPU to improve its speed. The simulation data and the actual data are compared for experimental verification, and the proposed algorithm is shown to effectively improve the quality of the ultrasonic echo image. When the echo image size is 153 Mbit, the operation speed is 16.29 times faster than that obtained by using the conventional synthetic-aperture algorithm. A processing speed of 6 frame·s-1 is achieved to satisfy the requirements of real-time processing for medical ultrasonic endoscopy.

Aiming at the problems of difficult feature extraction, poor classification accuracy, and less classification types in the remote sensing image multi-classification by the conventional methods, the feasibility of the convolutional neural network (CNN) model and the recognition effects of different CNN models are studied in the multi-classification recognition of hyperspectral remote sensing objects. The datasets are collected from Vaihingen provided by the international society for photogrammetry and remote sensing (ISPRS) and Google Earth. After the dataset-I containing six categories of ground objects is made, the dataset-II and dataset-III are made by adding ten and fourteen categories of ground objects, respectively. Through pre-processing image data, setting up network structures, adjusting model parameters, comparing network models, and so on, the classification accuracies of the above three datasets are all above 95%. By analyzing the influences of different CNN models on the multi-classification recognition of hyperspectral remote sensing objects, the feasibility and high recognition ability of CNN model in the multi-classification recognition of hyperspectral remote sensing are confirmed. The experimental results provide a certain reference for the application of CNN model in the multi-classification recognition of hyperspectral remote sensing objects.

A digital imaging method is proposed to measure the uncomfortable glare in interior lighting. Our study includes the systematic derivation of three parameters: the position index, solid angle from the viewer's perspective, and average luminance of glare source. To extract the profile information of the lighting sources, an adaptive threshold method was employed. The Unified Glare Rating (UGR) was calculated and analyzed with a digital camera image of actual office lighting conditions. The results show that the mean relative errors of the obtained position index, solid angle, and average luminance of the glare source were equal to 1.319%, 4.296%, and 3.001%, respectively. For a background luminance of 12.910 cd·m-2, the calculated value of the UGR was equal to 12.606, with a relative error of 4.326%. The advantages of this method are its simple calculations, high accuracy, and easy implementation. By applying our method to the glare test in indoor lighting, a theoretical framework for the development of a convenient and fast glare test system can be constructed.

Adaptive optics scanning laser ophthalmoscopy (AOSLO) has gained increased research attention over the past few decades. This study proposes a focus-adjustable AOSLO that uses a novel optical design comprising a Badal system, a deformable mirror, and a Shack-Hartmann wavefront sensor. This design corrects the low-order aberration of the human eye by setting a target and adjusting the distance between lenses in the Badal system. The optimum parameters of the Badal system based on the focusing and working principle of the system are analyzed; then, the system is simulated and optimized by Zemax software. Simulation results show that the spot size of the proposed system is smaller than the diffraction limit in the spot diagram. The Strehl ratio of any diopter is more than 0.8 and approaches as high as 0.95 for a low diopter. The modulation transfer function of every field approaches the diffraction limit. Moreover, the theoretical resolution in the normal human retina of this system (2.29 μm) is in a close range to the diffraction-limit resolution of the proposed system (2.11 μm). The proposed system can achieve a clear imaging of the human retina for people with a diopter from -6 m -1 to +6 m-1.

Scattering noise affects spectral data collected with spectral instruments. Spectral data curves measured for the same mineral species at different particle sizes and concentrations can produce an offset, which reduces the accuracy for matching the spectral data. To solve this problem, the present study reported a method based on the multi-scattering correction using merged augmented Lagrangian, in order to eliminate scattering noise and the resulting offset of spectral data; the method was firstly used accurate preprocessing, and then was used to similarity matching measurements combined with the spectral angles of the data. Six minerals and six pigments in the murals were selected as samples for the experiments. The spectral matching method was used to match and analyze the spectral data, which eliminated scattering noise and offset. Experimental results show that spectral data corrected using the proposed method are more accurately matched than uncorrected spectral data. In addition, the proposed method is more effective for mineral-species identification.

Coherent wind lidar uses the heterodyne detection method to amplify the backscatter signal by the local oscillation laser. Because the signal-to-noise ratio can reach the theoretical quantum limit, the lidar has high spatial and temporal resolution and high precision. Coherent wind lidar is widely used to measure wind shear, atmospheric turbulence, aircraft wake, gustiness, and gravity waves. Researchers in China and other countries are presently studying on coherent wind lidar. This study introduces the history of coherent wind lidar, describes the theory using different wavelengths, and briefly summarizes the development trends in this field.

At present, the efficiency and accuracy of face recognition based on visible spectra are strongly influenced by the factors such as lighting, pose, occlusion, expression change and photo fraud, especially the face recognition problems in the night vision environments need to be solved. A face feature extraction algorithm based on thermal infrared spectra is proposed. The thermal infrared face images are modeled to obtain a maximal data model, and the mixed model parameters are estimated and adjusted as a Gaussian mixture model. The isothermal features of thermal infrared Gaussian mixture face images are extracted, and the thermal feature images are reconstructed. The similarity between individuals measured by the calculation of probabilistic proximity index is used to present the recognition results. The experimental results based on the UCHThermalFace database show that the proposed method has relatively high recognition precision and robustness to multiple attitudes, feature changes, random occlusion and eye noise samples, which greatly improves the anti-interference ability of face recognition system in night vision environments.

Under normal temperature and pressure, the quantitative analysis of magnesium and titanium elements in aluminum alloy is carried out by handheld laser-induced breakdown spectroscopy (LIBS) to verify the instrument's stability and reliability. The handheld LIBS has a built-in 1064 nm Nd:YAG pulsed laser with a reflective coaxial structure. The AvaSpec-Mini 2048 spectrometer with a measurement wavelength range of 240-420 nm is used in the experiment. The characteristic lines of magnesium element at 285.17 nm and 383.83 nm and those of titanium element at 323.45 nm and 334.94 nm are selected. Then the calibration curve model is established with the partial least square method. The slopes of calibration curves for the same element under different characteristic lines are close to each other, and the values of linear correlation coefficients R are all above 0.9893. Moreover, the measurement results have good consistency with the standard values, and the relative error is no more than 8.70%. The detection limits of magnesium and titanium elements in the samples can reach the level of 10 -4. The handheld LIBS can be used to realize the quantitative analysis of magnesium and titanium elements in aluminum alloys and thus that the handheld LIBS has certain stable and reliable performances is verified, which is of great significance to the development of a portable LIBS.

The pink opal is a mineral aggregation of opal and palygorskite. The minerals composition and spectral characteristics of the opal have a significant difference with the traditional opal of play of color. The thermal variation behaviors of the pink opal are studied by means of differential thermal analysis and Fourier transform infrared spectroscopy (FTIR) under room temperature or variable temperature. The results show that the infrared absorption band of the pink opal characterizes the vibrational characteristics of silicate minerals palygorskite and silico oxide opal. The infrared absorption band caused by the structural water expansion vibration is mainly characterized by a group of infrared absorption bands with different intensities in the range of 4000-3000 cm-1. The infrared absorption band caused by sum frequency and frequency doubling of vibration frequency of structural water, crystalline water and adsorptive water is mainly concentrated in the range of 8000-4000 cm-1. With the increase of temperature, the relative absorption intensity of the absorption band is decreasing, and the absorption peaks also have different degrees of drift, splitting or addition. It shows that the binding of —OH has no relation to Si 4+, and is related to Mg2+, Al3+, Fe2+ and Fe3+ in palygorskite.

Aiming at the status of immaturity for the human-machine interaction technique of eye gaze tracking, a tabletop two-eye gaze tracking method is proposed based on pupil shape in space of stereo vision. With the low grey value distribution, the pupil center is located preliminarily. The radial derivative polar diagram in pupil area is used to extract the pupil edge point coordinates, and the random sample consensus (RANSAC) is used to fit the pupil edge with a suitable ellipse. The two-eye pupil edge point coordinates are matched using the ORB (Oriented brief) algorithm and the pupil edge point coordinates are obtained based on the two-eye stereo vision model. The least square method is finally adopted to calculate the pupil shape in space and the gaze direction is presented. The experimental results show that the positioning speed of pupil center is 300 frame/s, the two-eye gaze tracking speed is 15 frame/s, and the maximum gaze tracking error is 2.6°. It is verified that the proposed method has good accuracy, robustness and real-time performance, and it can be used in the field of human-machine interaction.