The existing method used to generate Gaussian noise signals is to generate random numbers through mathematical calculation. However, this method cannot achieve a true random signal nor can it match the actual noise signal. In this study, a Gaussian noise signal generator based on a quantum random number is proposed. A single-photon detector detects the photon signal of the selected path as the source of the random number and realizes the Gaussian noise signal generator based on a true random number. The random number is processed by a weighted Girvan-Newman Gaussian algorithm to obtain the Gaussian noise signal, which is implemented using the Verilog language in a field programmable gate array. The results show that the amplitude of the generated noise signal varies from 0 to 255. A statistical analysis of the amplitude spectrum obeys the Gaussian distribution. The power spectrum of the noise signal fluctuates uniformly at approximately 20 dB and follows a uniform distribution, thus satisfying the characteristics of the Gaussian white noise. Compared with existing methods based on quantum random numbers, the source of the random number is distinct in the present case, and the proposed generator can achieve real randomness. A simple scheme for realizing a true random number for a Gaussian noise signal generator is therefore presented.

To measure the two-dimensional velocity vector distribution of the flow field, a method of crossed tagging and displaying to conduct velocity vector measurement based on hydroxyl tagging velocimetry was developed. In this technique, the tagging laser beam and the displaying laser sheet are intersected to form tagging spots for flow tracing. Compared with the traditional cross-grid method to realize velocity vector measurement, this method uses a simplified experimental light path, which is advantageous to application in harsh environments. Moreover, the tagging spots are approximate dots, which are more convenient for identifying the tagging positions. Through the simulated tagging spot images, precise values of the signal noise ratio, spot size, and tagging shape were obtained. The velocity and fluctuation in a jet flame were obtained by a single tagging spot, and the velocity uncertainty can reach 2.2 m/s while an OH background is present in the flame. By using a cylindrical lens group to shape the tagging laser beam into multiple lines, 3×20 tagging spots were formed to measure the velocity vector distribution in the jet flame.

The direct-lit backlight used in an automobile head-up display must have high brightness and a small size. To reduce the thickness of a direct-lit backlight and improve its brightness, an optical brightness enhancement film is designed. First, a two-dimensional microstructure curve for the uniform illumination of light emitted by a single light-emitting diode (LED) is designed according to Snell's law. A microstructure curve is designed and calculated with an LED as an ideal point light source. Based on the consideration that the actual LED is a square luminous surface, the microstructure curve is optimized and its uniform illumination function to a real LED is improved. Then, based on the arrangement of LEDs, a hexagonal honeycomb arrangement is proposed and a simulation analysis is conducted. Simulation results show that the central brightness of use of the optical brightness enhancement film is increased by 173.2%. From the results of the microstructural design, a practical sample is produced by using the maskless direct lithography process and its effect is tested. Test results show that the central brightness is increased by 136.2% and the thickness is reduced by 13 mm. The proposed optical brightness enhancement film can effectively improve brightness and reduce the thickness of backlight units.

To obtain multi-degree-of-freedom micro-vibration information of an optical platform, a three-degree-of-freedom dynamic micro-vibration measurement system based on plane mirror dual-frequency laser interference is designed. The system utilizes the laser Doppler effect to obtain the three-point displacement information of the measuring mirror by using three measuring axes; calculates the overall displacement, torsion angle, and pitch angle information of the measuring mirror; and detects the three-degree-of-freedom micro-vibration in real time, thereby providing a basis for micro-vibration compensation control of the optical system. The uncertainty of the micro-angular vibration measurement system is analyzed, and an uncertainty model is established, which provides a theoretical basis for further improving the accuracy of the system and also lays a solid foundation for application of the system in the field of calibration and metrology. During the experiment, a high-precision seismometer was used to measure the platform simultaneously and, by comparison with the system translation measurement results, the measurement accuracy of the system was verified. The displacement resolution of the system is 5 nm, the torsion resolution is 5.05 rad, and the pitch resolution is 4.69 rad. The system has the advantages of multiple degrees of freedom, noncontact operability, high resolution, and traceability.

A beam scanning system is a fundamental component of a light detection and ranging (lidar) system, as it determines the field of view (FOV). To increase the FOV and maintain a specific angular resolution, multiple laser diodes are often used in a mobile lidar. However, the time response of different laser diodes is different, resulting in a mutual ranging difference between different ranging channels of the lidar. Risley prisms are outstanding beam pointing devices that offer good prospects for beam scanning of mobile lidars because of their notable features such as large FOVs and high pointing precision. In this study, the optical and mechanical configurations of a scanning system are designed based on the beam-pointing and scan-pattern models, which can realize two-dimensional scanning over a wide FOV. The horizontal and vertical FOVs of the scanning system are 360° and 30.4°, respectively. When the scanning system operates at a frequency of 5 Hz and scan lines of 30 with a 1-MHz repetition rate laser, the horizontal and vertical resolutions are 0.05° and 1.0°, respectively. These performance indicators are comparable with those of mainstream mobile lidars and can meet the actual requirements of mobile lidars. Most importantly, this scanning system can eliminate ranging differences between different channels, and significantly improve the ranging precision of the lidar.

To improve the optical properties and laser damage threshold of TiO2 films and to ensure their wider use and greater stability in lasers, TiO2 films were deposited on K9 glass using electron beam evaporation technology, and the samples were post-treated with Ar/O mixed plasma. The effects of plasma post-treatment time on the laser damage characteristics of TiO2 thin films were analyzed by measuring the film spectral characteristics, surface defect density, surface topography, damage threshold, and damage topography. The experimental results showed that the refractive index and fill density of the films increased with increasing post-treatment time, whereas the physical thickness and surface roughness decreased. In addition, the surface defect density first decreased and then increased. After a 20-min post-treatment, the LIDT of the TiO2 film samples increased from 5.6 to 9.65 J/cm2 under a 1 064-nm laser irradiation, representing an increase of 72.3%.

To measure the angle shift that occurs under high-temperature conditions, a fiber-based high-temperature-resistant incremental encoder was developed in this study. First, a high-temperature-resistant fiber was employed to transfer light under high-temperature conditions. A field programmable gate array was used to process the incremental encoder's electric signal, and a USB was used to transfer the angle in a normal temperature environment. Second, the effect of the collimation degree of a light beam on the performance of the optical encoder was analyzed, and then a light-emitting diode (LED) convex lens fiber couple was developed. Finally, 23 surface polyhedrons and autocollimations were used to test the accuracy, and the accuracies of the LED convex lens fiber couple and fiber direct couple were compared. The results show that this system can work normally in 100 ℃ conditions. The size of the encoder head is 62 mm× 42 mm, the resolution is 0.3", and the precision 3σ is 13.55 ″, which represent an improvement of 20.8% over that of the fiber direct couple. The system thus satisfies high-temperature conditions.

To address the urgent need for ultra-large aperture and lightweight high-resolution optical payloads in the field of high-orbit military monitoring, an image inversion and quality enhancement method was proposed for large-caliber diffractive imaging systems with diffractive optical elements used as the primary mirror. First, an image inversion model was established for the diffraction imaging system based on the regularization and iso-halo block theories. Here, the multi-regularization and multi-constraint algorithm was adopted for theoretical model optimization, and two bottleneck problems were in turn solved to a certain extent. One was the image blurring caused by both the large-caliber optical system aberration and the diffraction effect of the main mirror; the other was the large spatial variation of the point spread function. Nonlinear transformation and wavelet threshold filtering were then used to formulate a contrast enhancement method to address the low signal-to-noise ratio and low contrast problems that derive from low diffraction efficiency and non-design level sub-background clutter. Finally, experiments and a simulation were conducted to test the proposed method. Experimental results show that the structural similarity of degraded and original images exceeds 08, the Signal to Noise Ratio(SNR) is increased by 10% and fidelity is more than 80% with CLE of 30. The proposed method improves image quality, including image clarity and contrast as well as the suppression of noise and background radiation. The proposed method thus has great theoretical significance and engineering application value for the practical application of ultra-large aperture thin-film diffraction systems.

To improve the Graphical Representation of Model Spaces (GRMS) index of optical elements, a new polishing method based on smoothing surface shape by fitting was proposed after the rules of the effects of surface shape precision and distribution on the wavefront gradient index were summarized. The fundamentals and workflow of this method were then presented. First, the surface shape discontinuity was determinedby scanning and computing the original surface shape data. Then, the new surface shape data used in guiding the processing were generated using the non-uniform rational B-spline fitting algorithm and the data between the neighbor discontinuities was reconstucted. Finally, appropriate processing parameters were established based on the generated surface shape data. An experimental study was conducted on K9 material plane mirrors of 610 × 440 mm, and results showed that the proposed method can reduce the GRMS from 11 nm/cm to less than 7.7 nm/cm in two polishing cycles while maintaining a high surface shape precision.

Small-diameter shape sensors have a variety of potential applications in medicine such as the positioning of breast tumor chest needles, shape display of intestinal endoscopes, and positioning of cardiac vascular catheters. The shape-reconstruction and end-positioning accuracies of shape sensors have always been major concerns of researchers. To improve the shape-reconstruction accuracy of a fiber Bragg grating (FBG) shape sensor, the strain transfer law of an FBG shape sensor with 90° distribution of fiber around the center of the substrate was studied and verified experimentally. Epoxy resin glues the FBG to the surface of the nickel-titanium alloy wire to form a four-layer structure of the substrate-bonding layer-fiber-bonding layer of the FBG shape sensor. Based on a simplified model of strain transmission mechanics of the FBG sensor, the corresponding average strain transfer rate formula was derived. The effects of the fiber-to-substrate center distance, bond layer length, thickness, and elastic modulus on the average strain transfer rate of FBG sensors were also analyzed and experimentally verified. Experimental results show that the average strain transmission rate of the packaged FBG sensor increases with increasing substrate length in the effective range, whereas an increase in the outer diameter of the bond layer has little effect on the average strain transmission rate, thus verifying the applicability of the theoretical model. The strain transfer rate is introduced into the shape reconstruction, and the shape-reconstruction accuracy of the FBG shape sensor is reduced from 3.5% to 2.7%.

To measure the center of gravity of a large-sized aircraft, a flexible center of gravity measurement method is proposed. A method to obtain the line that passes through the center of gravity, the coordinate transformation, and the means to obtain the center of gravity itself are also proposed. First, the projection of the line that passes through the center of gravity is measured, and a line that passes through this projection point is established and converted into a coordinate system of the body under test. Second, by means of rotating or tilting the body at any angle, the second line using the same method is measured and converted into the coordinate system of the body. The center of gravity is obtained by finding the intersection of the two lines. Finally, a Monte Carlo simulation analysis method is used to determine the method's uncertainty. Both the simulation analysis and experimental results showed that the center of gravity measurement error is within 0.3 mm. Thus, the proposed method can achieve high-precision detection of the center of gravity even when the position between the device under test and the measuring device is unknown.

Acoustic emission is the most direct and sensitive characterization of grinding interference. Therefore, to ensure high-precision grinding, acoustic emission signals were used to evaluate the degradation in grinding performance of a diamond wheel. A fused silica grinding experiment was conducted using a newly trimmed diamond grinding wheel to record the acoustic emission signal and surface morphology of the grinding wheel during its life cycle. A wavelet packet analysis was used to determine the typical low-frequency band of the abrasive wear of the grinding wheel. The frequency energy ratio was then extracted by using principal component analysis to characterize the degradation of grinding wheel grinding performance. The results showed that the acoustic emission frequency energy ratio was completely independent of the processing parameters. As the abrasive wear of the grinding wheel intensified, both the proportion of large-scale rupture and the amplitude of the corresponding low-frequency characteristic increased. When the abrasive particles were worn to the limit, the abrasive grain was broken to produce a new cutting edge, improving the removal ability of the grinding wheel, but the variance of the spectrum was significantly increased, inhibiting the wheel from maintaining a stable quality in ultra-precision machining. The variation of the white pixel ratio of the grinding wheel shape image verified the correctness of the acoustic emission signal analysis results.

To implement low-cost mass production of large-area micro/nanostructures on large-size rigid and fragile substrates, this study presents a novel composite nanoimprint lithography technique. First, the basic principle and process of the composite imprint lithography are described. Then, the effects and rules of the key process parameters (i.e., spreading speed of the flexible mold, imprint force, imprint speed, and curing time) on the imprinted patterns are revealed by a series of experiments. Finally, through a combination of the home-made composite imprinter and optimized process parameters, four typical cases are illustrated. We fabricated large-area micro/nanostructures on three rigid substrates (glass, polymethyl methacrylate, and sapphire), including microscale columnar structures (the largest graphic area is 132 mm× 119 mm), a microscale grating structure (the largest area is 6-inch round), and a nanoscale columnar structure (the graphic area is 47 mm× 47 mm). Experimental results show that the presented composite imprint lithography technique provides a novel way for the macroscopic quantity preparation of large-area micro/nanostructures and large-area patterning on rigid or fragile substrates, thus presenting a bright prospect for industrial applications.

Few real-time methods exist for three-axis turntable tracking when attemping to solve the orientation change of targets. The alt-azimuth and alt-alt methods, which lock the roll and azimuth axes of the turntable, respectively, are mostly utilized. However, these two methods suffer from the problem of tracking in blind zones. In this study, a minimum deviation of triaxial velocity (MDTV) algorithm was proposed to improve the performance of three-axis turntable tracking. A motion model of a three-axis turntable was first established, and the minimum deviation of triaxial velocity was used as the index. The motion of the target, angular position, and angular velocity were then referenced. Finally, a general inverse matrix of the motion model was deduced. The results indicate that triaxial deflections are uniquely determined for distributing the orientation changes of a target into the triaxial tracking evenly, achieving a real-time servo in the three-axis turntable. Compared to the alt-azimuth and alt-alt tracking methods, the MDTV method significantly reduces the angular velocities and accelerations in the tracking process and ensured smooth running when tracking targets are in a blind zone. When a rotating target is tracked, the maximum pointing deviation with the MDTV method is only 15.4% of that with the alt-azimuth method. The study thus proves that the MDTV method can solve the problem of tracking in blind zones and improve the tracking accuracy of a three-axis turntable.

To address the deficiencies of the traditional cross-coupling control structure, an active compensation control structure was proposed in this study to reduce the contour error of an X-Y linear motor precision motion platform. First, a permanent magnet synchronous linear motor was used to analyze the single-axis positioning tracking error, where the tracking error was related to the position reference. Therefore, adjustment to the amount of coupling compensation to the reference instruction was added to the system, which considered the disturbance of the reference command in the actual working condition. The contour error components were compensated to the position and speed loops of each axis servo and verified by a simulation and an experiment. Results show that the average value of the contour error of the X-Y axis motion platform when tracking a large curvature complex track using an active compensation method is 20.68 μm. The single-axis tracking error is 70 μm. Compared with the traditional cross-coupling control structure, the accuracy of the active compensation structure contour error is improved by 15.5%, the single-axis tracking error is reduced, and the reference command disturbance can be suppressed.

A repetitive locking/unlocking mechanism based on a Shape Memory Alloy (SMA) wire was proposed to meet the multiple locking/unlocking requirements of the servo system of an onboard laser communication terminal during orbit. First, the working principle of a repeated locking/unlocking mechanism that used a spring component to realize locking and that of an SMA wire output return stress to unlock were analyzed. The locking principle was applied to optimize the locking force and displacement. Second, the locking force was simulated and analyzed. The balance between the unlocking force and the influence of the locking force on the modal of the whole machine was also considered. Finally, a prototype was developed, and the repeated locking/unlocking function as well as the relationships among the unlocking force, unlocking time, and supply current were verified experimentally. Problems such as high temperature, vibration, and different aerospace environments that affect the reliability of the locking mechanism were investigated. Experimental results show that the mechanism can provide a 1 000-N locking force. When a 12 V/3 A power supply is input, the maximum unlocking time does not exceed 2 s. Under an ambient temperature of 90 °C, stable and reliable locking can be ensured, and multiple repetitions of locking/unlocking can be realized. The mechanism has the advantages of being light, small, and repeatable and of having low impact and low power consumption, thus meeting the requirements of multiple locking/unlocking of an onboard laser communication terminal. The mechanism thus demonstrates great application potential.

To analyze the effects of radial temperature gradients on the surface figures of lightweight reflective mirrors, a 2-m-aperture lightweight mirror of a space camera was taken as the research object and a finite element model of the mirror was established. The effects of different radial temperature gradients and lightweight structures on the surface figure Root Mean Square (RMS) values of reflective mirrors was analyzed, and the accuracies of a simulation method and conclusions were verified through an experiment. In addition, the coupling effects of different lightweight structures and radial temperature gradients on the mirror surface figure were studied. Results show that the surface figure RMS values derived from the radial temperature gradients of different distribution modes can differ by a maximum of 294 times, and the surface shape error caused by the radial gradient is difficult to reduce when optimizing the mirror support scheme. The mirror surface figures of different lightweight structures are different for the radial temperature gradient sensitivities of different distribution modes. The lightweight methods of trefoil and symmetric chamfering are thermally stable to the mirror in a specific radial temperature field distribution mode. Finally, sensitivity has an adverse effect.

To solve the problem of excessive random vibration response of a micro-scale mechanical-sensitive component of an optical instrument for a space station, micro-scale and fine modeling of the component was conducted, its mechanical properties were studied, and effective measures were adopted to suppress its random response. First, the boundary coordination of the micro-scale model and the variation problem of the micro-scale finite element were discussed. On this basis and according to the characteristics of the micro-scale structure, four models were established. A system-level random vibration simulation was performed, and the characteristics of the four micro-scale models were studied. Following analysis and comparison of the random response, including the acceleration, displacement, and stress, the micro-scale modeling method was discussed in-depth. Then, random vibration response suppression and a simulation were separately conducted. Finally, mechanical and thermal tests and a testing experiment were performed. Results reveal that the optical instrument has normal functions, the root mean square of the acceleration response of the micro-scale structure is reduced by 42.4%, the stress is reduced by more than 20%, and the stress safety margin is far greater than zero. The maximum relative error of acceleration response for the simulation and test is within 10%. The study thus showes that the proposed four-element modeling method for a cross-section of a micro-scale structure is accurate and reliable.

To solve the difficulty of ensuring the shape accuracy of micro grooves in a micro device, a new method of using multi-station shape-angle electrodes through Electrical Discharge Machining (EDM) was proposed. A total of 304 stainless steel samples were used as workpiece materials, and pure copper and copper-tungsten alloy (W75%) were selected and compared as electrode materials. To reduce the influence of electrode wear and installation error, multi-station shape-angle electrodes were machined by low-speed Wire-cut EDM (WEDM), and then a 16° and 0.36-mm-depth groove was formed by EDM. Results show that a multi-station shape-angle electrode having enhanced surface quality can be fabricated using a lower feed rate, a constant speed, and multiple cutting through WEDM. The forming accuracy of the grooves was improved with the increase in the number of stations, and the forming accuracy of the copper-tungsten alloy was improved as a result of its low loss at the same station number. Therefore, a six-station copper-tungsten electrode was used to machine the workpiece with a sharp angle of 16.3° (relative error of 1.9%), a 0.356-mm depth (relative error of 1.1%), and improved form accuracy. This method can meet the processing accuracy and application requirements of sharp grooves in a micro device.

To realize the intelligent diagnosis of bearing faults, an intelligent fault diagnosis method for the thin-wall bearing of a robot based on information fusion was studied. First, a test and multi-information data acquisition system of the thin-wall bearing of a robot was built by acquiring acoustic emission and vibration acceleration signals. Then, data from acoustic emission and vibration acceleration signals detected during the test of thin-wall bearing under different fault types, equivalent loads, and rotational speeds were obtained using an orthogonal experimental method. A thin-wall single-row angular contact ball bearing (ZR71820) was used as the research object, and pitting and micro-crack defects were produced on the bearing outer ring, inner ring, and rolling bod. Finally, the root mean square value and kurtosis index in the time domain, as well as the root mean square frequency in the frequency domain, were selected as the characteristic parameters of the vibration and acoustic emission signals. Fault diagnosis of thin-wall bearings based on single vibration or acoustic emission signals was conducted. In addition, an intelligent fault diagnosis of thin-wall bearings were researched based on the fusion characteristics of acoustic emission and vibration acceleration signals using Self-Organization feature Map (SOM) and Back-Propagation (BP) neural networks. Experimental results indicate that the accuracies of fault diagnoses based on vibration signals, acoustic emission signals, and BP and SOM neural network information fusion are 85.7%, 81.0%, 93.5%, and 95.2%, respectively. The accuracy of intelligent fault diagnosis based on SOM neural network information fusion of the thin-wall bearing is 9.5%, 14.2%, and 1.7% higher than that of single vibration, acoustic emission signals, and BP neural network information fusion, respectively.

To achieve fast and accurate segmentation of natural images with intensity inhomogeneity and complicated backgrounds, an active contour model combined with Retinex correction and saliency analysis for image segmentation was proposed. Retinex correction was applied to obtain the reflection component of objects in images; this could suppress the influence of intensity inhomogeneity caused by nonuniform illumination. Moreover, the Retinex-corrected image reflected the image essence more objectively, ensuring the accuracy of subsequent salient information extraction and making it more practical and instructive. The introduction of saliency information into the active contour model was helpful for the effective segmentation of images with complex backgrounds. By combining Retinex correction and saliency information, a new active contour model energy equation was obtained, and the level set method was used to guide the curve evolution to achieve image segmentation. Through experimental analysis, the proposed method was proved to be fast, effective, and robust. The average processing time on the MSRA database is 4.277 s per image, and the average F value is 0.735.

In order to overcome the problem that existing airborne methods for LiDAR point cloud classification have difficulties in obtaining high classification accuracy and reducing processing time simultaneously, a method using transfer learning for classifying airborne LiDAR point clouds was proposed. First, normalized height, intensity, and normal vector were calculated for each LiDAR point, by setting different sizes of neighborhood, and multi-scale point cloud feature images were generated by utilizing the proposed feature image generation strategy. Subsequently, a pre-trained deep residual network was employed to extract multi-scale deep features from the generated multi-scale feature images. Finally, a neural network model containing only two fully connected layers was constructed to achieve efficient training, and each point cloud was classified by the trained neural network model. Experimental results of two ISPRS benchmark airborne LiDAR point cloud sets demonstrat that the proposed method requires less training time, and the overall classification accuracy obtained by the method is 89.6%, which is 4.4% higher than the best classification result reported on the ISPRS official website. The classification result can provide reliable information for further processing and application of point cloud.

To improve the visualization effect and enhance the local detail information of an image, a tone-mapping method that combines guided filtering and nonlinear second-order features was proposed. First, the luminance information of an input image was extracted and the luminance image was decomposed by a multi-scale using a guide filter to obtain a base layer and a detail layer image. Second, a weighting map of the detail layer was constructed by the guided filtering method, a nonlinear second-order feature was constructed using a Hessian matrix for the basic layer, and a weight map of the basic layer was constructed using the guided filtering method. Finally, the basic graph was decomposed according to the weight map. The layer and detail layer images implement reconstruction of the luminance image and then restore the color information of the luminance image to obtain the final resulting image. Experimental results show that the method preserves the local detail information of the source image and has a good visual effect. In this study, the Hessian matrix is used to extract the high-frequency information of the base-layer image, which can more effectively highlight the edge information and enrich the detail features of the image. Compared with the objective index of the comparison algorithm, the quality score of the method, the structural fidelity, and the natural similarity are increased by 11.28%, 10.82%, and 186.46%, respectively.

To solve the insufficient tracking capability problem for a fully convolutional Siamese network (SiamFC) in complex scenarios such as those involving fast motion and large similar interference, SINT was introduced as a redetection network to improve the SiamFC. When multiple peaks appeared in the tracking response map, the proposed algorithm enabled the redetection network to redetermine the target position with higher accuracy. At the same time, a generative model was adopted to construct a template to adapt to various appearance changes of the target, and a high-confidence model update strategy was used to avoid the model corruption problem. Our algorithm is tested on OTB2013, and nine state-of-the-art algorithms are selected for comparison. The tracking accuracy of our algorithm reaches 88.8%, the best among all the algorithms selectes for comparison, and the success rate reaches 63.2%, which is the second best. Both these properties offer considerable improvement over the SiamFC results. Analysis of several representative video sequences demonstrate that our algorithm has high accuracy and strong robustness in cases involving fast motion, severe occlusion, background clutter, illumination changes, and long-term tracking.

Layout optimization of inspection points is key for the inspection of geometric errors in models with complex surfaces. To deal with difficulties in the adaptive planning of whole inspection points of model components that were characterized by multiple surface characteristics, mesh-based adaptive planning of inspection points was studied. A complex surface model was transformed into a dense triangular mesh model, and the triangular mesh simplification method was used for adaptive planning of inspection points. The local polynomial fitting and local curvature estimation technology were used for adaptive reduction of mesh vertexes. The vertex substitution method was used to avoid any influence from the meshing discretization error. The subset selection technology was used to improve reduction efficiency. The experimental results show that this method should not introduce the meshing discretization error of the surface, and that it was not sensitive to meshing granularity. The maximal error of the mesh-based method described here should reduce by 32.8% under the same detection points in comparison to that of the uniform sampling method, and should reduce by 16.9% in comparison to that of the random Hammersely sequence method. The mean error of the mesh-based method should reduce by 28.7% and 18.5% in comparison to that of the uniform sampling method and the random Hammersely sequence method, respectively. Meshing can avoid coordination difficulties among multiple surfaces, and the planned detection points can objectively reflect the processing quality of the complex surface.

In this article, an extraction method of line contour points from point cloud data (for a fold-line edge and a ridge edge with a small chamfer) is proposed. In this method, multiple-direction slices were constructed to obtain section data containing line contour points. Then, according to the form characteristic of the section data, a pattern vector was constructed by using the normal angle of two adjacent points to represent the characteristic. The section data was divided into many small groups, each with the same number of components. Euclidean distances between the pattern vector and each small group were calculated, and the group with the smallest Euclidean distance was located for picking out the line contour point. To determine the accuracy of the proposed method, examples of three kinds of edges were used: a straight-line edge, a curve edge, and four circles. The extracted line contour points were fitted to a straight line, a three-order polynomial, and circles, by the least square fitting method. The deviations of the line contour points to the fitted lines were calculated. The standard deviations of the straight-line and curve edges were 0.076 and 0.047 mm, respectively. The standard deviations of the four circles were smaller than 0.1 mm. The radius deviation of the four circles to the reference data, measured by a coordinate measuring machine, was within 0.1 mm. This pattern vector method is applicable for extracting line contour points on a fold-line edge and a ridge edge with a small chamfer. When the scanning accuracy is higher than 0.03 mm, the pattern vector method ensures an accuracy of 0.1 mm.

To solve the problems of high calculation amounts and low classification accuracy of Gaussian process regression in hyperspectral image classification, a spectral-spatial joint classification algorithm for hyperspectral images based on improved Gaussian process regression was proposed. A subset of samples was selected using maximum variance as the index to narrow the calculation range of the Gaussian process regression parameter solution, and a square root matrix decomposition method was introduced to predict the model results for incoming added samples, all of which effectively improve the efficiency of calculation. A spatial-spectral correlation distance of neighborhood pixels was redefined in the pixel neighbor space based on spatial-spectral feature information. In addition, a space-spectrum correlation distance integrated with spatial neighbor information was used as the weight to measure the similarity of neighborhood pixels. These increase the probability that similar features would be classified as neighbors, thus improving the accuracy of feature classification. Simulation experiments were conducted on two sets of hyperspectral datasets from Indian Pines and Pavia University. Experimental results show that, compared with other similar algorithms, the proposed algorithm improves overall classification accuracy, average classification accuracy, and the Kappa coefficient by at least 2.3%, 1.4%, and 1.07%, respectively. Compared with the model algorithm prior to enhancements, the improved algorithm not only achieves higher overall classification accuracy but also considerably reduces the running time.

Current fatigue monitoring equipment is expensive, non-portable, and its algorithm has a low robustness in eye movement image processing. Existing front-end camera fatigue monitoring schemes have many limitations and greatly affect their users efficiency; as such, the development of portable and accurate fatigue monitoring systems is very challenging. In this paper, a new type of fatigue monitoring glasses and N-Range image processing algorithm is proposed to improve the robustness of eye location analysis and accuracy of fatigue detection. Real-time fatigue analysis is performed for users of the proposed mechanism according to the theory of percentage eye closure (PERCLOS) P80 fatigue assessment; it is considered that the P80 criterion is the most suitable for this study. In the process of side-eye image processing, an N-Range eye region extraction algorithm is proposed. The activation map is computed by an N*N convolution kernel, and it is segmented by the OTSU threshold segmentation method. The activation value of pixels smaller than the threshold is set at zero. Based on this, the standard deviation projection in the horizontal and vertical directions is calculated; the human eye region is located by the average threshold method from the projection map. The eye closure degree is measured by calculating the ratio of eye height to eye width and then counting the closure time. Experiments show that the aforementioned problems can be adequately solved by our method. Even in complex environments, the method in this paper still performs well, with fatigue judgment accuracy reaching 94%. The fatigue-monitoring scheme proposed in this paper, can be positively adapted under many uncertainties. As such, our fatigue monitoring glasses and N-Range algorithm can achieve high accuracy without affecting the efficiency of workers.