Existing representation learning methods of cultural relics require numerous labels. Manual labeling is time-consuming and labor-intensive. Furthermore， supervised learning methods cannot effectively learn the internal structure information of point clouds. We propose an unsupervised representation learning network to extract the deep features of ceramic cultural relics. The approach is based on local-global bidirectional reasoning. First， we propose a multi-scale shell convolution-based hierarchical encoder to extract local features at different scales. Second， the local-to-global reasoning module is used to map the extracted local features to the global features. The differences between the two types of features are measured using metric learning for iterative learning. Third， a fold-based decoder is used to obtain better reconstruction effects from the acquired global features in a coarse-to-fine manner. A local-to-global reasoning module supervises only the local representation to be near the global one. We propose using a low-level generation task as a self-supervision signal. The global feature can capture more basic structural information about point clouds， and the bidirectional inference between local structures and global shapes at different levels was used to learn point cloud representations. Finally， the learned representations are applied in the downstream task of point cloud classification. Experiments on the Terracotta Warriors and ModelNet40 datasets show that the proposed model significantly improves in terms of classification accuracy. The classification accuracies were 93.33% and 92.02%， respectively. The algorithm improved by approximately 4.4% and 2.82% compared with the supervised algorithm PointNet. The results demonstrate that our model achieves a comparable performance and narrows the gap between unsupervised and supervised learning approaches in downstream object classification tasks.

To solve the loss of detail information and insufficient feature extraction in the fusion results of infrared and visible light images， a deep learning network model for infrared and visible light image fusion with multi-scale densely connected attention is proposed. First， multi-scale convolution is designed to extract information of different scales in infrared and visible light images to increase the feature extraction range in the receptive field and overcome the problem of insufficient feature extraction at a single scale. Then， feature extraction is enhanced through a densely connected network， and an attention mechanism is introduced at the end of the encoding sub-network to closely connect the global context information and enhance the ability to focus on important feature information in infrared and visible light images. Finally， the fully convolutional layers that compose the decoding network are used to reconstruct the fused image. This study selects six objective evaluation indicators of image fusion， and the fusion experiments conducted on public infrared and visible light image datasets show that the proposed algorithm exhibits improved results compared with eight other methods. The structural similarity （SSIM）， spatial frequency （SF） indicators increase by an average of 0.26 and 0.45 times， respectively. The fusion results of the proposed method retain clearer edge and target information with better contrast and clarity， and are superior to the compared methods in both subjective and objective evaluations.

Targeting negative effects such as clarity and contrast degradation and color distortion of images acquired in hazy weather， underwater， and in nighttime environments， a two-stage image restoration method using an improved atmospheric scattering model is proposed. A global compensation coefficient is introduced into the traditional atmospheric scattering model to obtain an improved atmospheric scattering model； the two-stage image restoration method based on this model consists of two stages. First， a degraded image is fed to the improved atmospheric scattering model to obtain a coarse restored image. The grayscale world algorithm is then used to determine the albedo of this coarse restored image. Second， the albedo and output image of the first stage are fed to the improved atmospheric scattering model to obtain the final restored image. Experimental results indicate that the proposed method can avoid the problems of color distortion and dark tones in the restored images and has good applicability. The method can effectively achieve image dehazing， underwater image restoration， and night image enhancement. The proposed method achieves excellent results in both quantitative and qualitative experiments compared with state-of-the-art methods.

Lens-less imaging is affected by twinning noise occurring in in-line holograms， and the reconstructed results continuously face poor reconstruction signal-to-noise ratio and low imaging resolution. This study proposes a lens-less imaging via a score-based generation model. In the training phase， the proposed model perturbs data distribution by gradually adding Gaussian noise by using a continuous stochastic differential equation （SDE）. A continuous time-dependent score-based function with denoising score matching is then trained and used to solve the inverse SDE required to generate object sample data. In the testing phase， a single Fresnel zone aperture is used as a mask to achieve lens-less encoding modulation under incoherent illumination. The prediction-correction method is then used to alternate iteration steps between the numerical SDE solver and data-fidelity term to achieve lens-less imaging reconstruction. Validation results on LSUN-bedroom and LSUN-church datasets show that the proposed algorithm can effectively eliminate twin image noise， and the peak signal-to-noise ratio （PSNR） and structural similarity （SSIM） of the reconstruction results can reach 25.23 dB and 0.65， respectively. The PSNR values of the reconstruction results are 17.49 dB and 7.16 dB， which is higher than that of lens-less imaging algorithms based on traditional back propagation or compressed sensing， respectively. In addition， the corresponding SSIM values were 0.42 and 0.35 higher， respectively. Therefore， the reconstruction quality of the lens-less imaging is effectively improved.

Next generation large-scale spectroscopic survey projects may require smaller and high-precision fiber-positioning units. In this paper， a design scheme of the precise positioning part of the R mechanism of an R-θ fiber-positioning unit is proposed. Stacked piezoelectric ceramics combined with a flexible hinge lever displacement amplifying mechanism are used to achieve positional accuracy of the precise positioning part of the R mechanism. To minimize the influence of hysteresis nonlinearity on the precision of precise positioning， the classical Preisach model is used to establish the driving model of the amplified output displacement， and the corresponding optimization is conducted on this basis. The average error of the final established precise positioning displacement model is within 4 μm， and this error is significantly reduced compared with the maximum nonlinear error of 40 μm. The experimental results reveal that the displacement-driven model established by the Preisach model further improves the positioning accuracy and meets the error design requirements of precision positioning and also allows for a larger design space for later coarse positioning.

To address the finishing difficulties of existing medical titanium alloy bone screws， such as complex structure， different sizes， and surface burrs， a new finishing method known as drum type ultrasonic magnetorheological compound finishing was proposed， and a drum-type ultrasonic magnetorheological compound finishing device was designed independently. The device adopts an electromagnetic excitation mode with a three-magnetic pole excitation structure. The structural size and coil parameters of the excitation device are determined by theoretical calculations and a Maxwell simulation analysis. A roller-type ultrasonic–magnetorheological compound finishing experimental platform is built. An excitation current single-factor polishing experiment is performed. The magnetic induction intensity of the MR fluid in the finishing device is measured and compared with the simulation results. When the width of the n-pole of the electromagnetic excitation device is 35 mm， the number of turns of the coil is 1 080， and the wire diameter is 1.25 mm， the maximum current that can pass through the coil is 5 A. Moreover， the finishing area can form a good magnetic circuit， the magnetic field distribution can form the magnetorheological and abrasive finishing areas， and the magnetic induction intensity of the magnetorheological finishing area can reach up to 0.57 T. The magnetic induction intensity of the MR fluid in the finishing area of the drum is less than the simulation value. The surface roughness change rate of the workpiece before and after polishing increases first and then decreases with an increase in the excitation current. When the current is 4 A， the surface roughness decreases from 1.39 μm to 0.435 μm， and the maximum surface roughness change rate is 68.7%. The magnetic induction intensity of the electromagnetic excitation device with a three-pole excitation structure can meet the requirements of the magnetic field in the finishing of MR fluids， and the effective polishing area is large.

Electron beam direct writing （EBDW） technology is an ideal tool for fabricating micro curved-surface electronics， which has high resolution and simple operation. The absorption energy deposition density distribution of resist directly affects the accuracy and resolution of the exposure pattern， but the existing plane process is no longer suitable for curved-surface direct writing because of its asymmetric distribution. In this paper， Monte Carlo simulation based on micro-cube element is used to calculate the absorption energy deposition density distribution under different direct writing parameters. The simulation results are shown that the ellipticity of the exposure dot increase with increasing incident energy or with the incident tilt angle increases. By reducing the beam spot size and thin layer， the resolution by direct writing on the curved can be improved. The experiment results are shown that aspect rations of the exposure dot are 1.458，2.323， 2.924， as well as 1.014，1.113，1.173 with incident energy （5 keV， 10 keV， and 15 keV） and incident angle （5°，10°， and 15°） parameters respectively. The increased incident energy on the ellipticity is even more obviously. The results of this study provide a theoretical basis for practical direct writing on curved-surface and are of relevance to the process.

Bioluminescence tomography （BLT） is a promising in vivo molecular imaging tool that allows non-invasive monitoring of physiological and pathological processes at the cellular and molecular levels. The reconstruction accuracy of BLT is affected by the optical transmission model error and the ill-posedness of the inverse problem. The higher-order optical transmission model was able to improve the precision， while the multi-spectral method was able to alleviate the ill-posedness of the inverse problem. In this study， the spectral differential strategy， combining the spectral differential theory and multi-spectral method， was applied on the optical transmission model based on the diffusion approximation equation （DE） and third simplified spherical harmonic approximation equation （SP3）. First， errors in these two radiative transfer equations （RTE） approximations were analyzed， and the attenuation effect on the error was compared when the spectral differential strategy was applied on two types of optical transmission models. The forward simulation experiment results showed that the spectral differential strategy can effectively reduce the model error of the DE and SP3 models. The spectral differential strategy resulted in the transmission accuracy of the DE model resembling that of the SP3 model， and decreased the high requirements on computing time and storage space of high-order approximation. On this basis， the spectral differential strategy was applied on the DE and SP3 optical transmission models for light source reconstruction. The experimental results showed that the spectral differential strategy not only improves the accuracy of the two light transmission models， but also alleviates the ill-condition of the inverse problem in BLT， yields a location error of reconstructed source within 1 mm， and improves the accuracy of the light source reconstruction in target location， shape restoration， and image contrast. The average time required by the SP3 model was approximately 1 525 s. In contrast， the DE model combined with spectral differential strategy had an average time consumption of only approximately 34 s， resulting in balanced reconstruction accuracy and speed.

As the number of interconnected layers on the front side of integrated circuits increase， it becomes increasingly difficult to detect the internal electrical signals from the front side of integrated circuits. A probing optical path based on the common-path interferometer is designed in this study to use a laser for detecting the internal electrical signal waveform from the back side of an integrated circuit without making contact. To this end， a signal processing method based on a lock-in amplifier is proposed to extract the weak electro-optical signal of the device carried in the reflected light， which exploits the strong noise rejection and high sensitivity of the lock-in amplifier technique. Considering that the electro-optical signal of the device varies periodically with the electrical signal， the electrical signal from the photodetector is first processed by the lock-in amplifier to eliminate most of the noise， and then the averaging technique is employed to further suppress the external noise and improve the signal-to-noise ratio of the electro-optical signal of the device. Finally， the electro-optical signal of a device drowned in noise is extracted and the electrical information inside the device is reconstructed. The electro-optical signal of the circuit node with a dynamic operating current in the order of μA inside the chip is successfully detected using the proposed method. The signal-to-noise ratio of the extracted signal reaches 4.99 dB， while the signal-to-noise ratio of the signal obtained solely via the averaging technique is only -44.29 dB. This paper presents a novel method for the optical detection of electrical signals inside integrated circuits， which can be applied to perform dynamic defect detection of integrated circuits in the future.

In order to obtain the evolutionary process and flow-field shape of shock waves emanating from an explosion wave simulation device outlet， a visualization research of explosion shock waves is conducted. This research can aid in better control of the loading waveform， and consequently， meet the loading requirements of weapons. A reflective high-speed schlieren measurement system is designed， which comprises a high-speed charge-coupled device （CCD） imaging unit and a series of optical elements. The image-evolution process of explosion waves can be measured and compared with the signals received by an external shock wave pressure sensor. The research shows that the number of explosion shock wave fronts obtained by the schlieren system is consistent with the results measured by the shock wave sensor， and the image morphology， formation， and evolution process of shock and reflected waves are obtained. Concurrently， the velocity information of explosion shock waves is obtained through image data of multiple shock wave fronts. The results provide a new technical approach and basis for analyzing the effective loading of explosion waves， as well as a better understanding of the form and evolutionary law of shock wave pressure formation.

Grating projection three-dimensional measurement technology has become one of the most widely used three-dimensional measurement technologies owing to its noncontact， high resolution， and high accuracy advantages. However， for objects with highly reflective surfaces， the collected image appears partially too bright or dark， which results in point cloud measurement loss， thereby further affecting the measurement accuracy. To accurately measure the three-dimensional contour of objects on high reflective surfaces， an improved multi-exposure image fusion technique is proposed. Images of different exposure durations under white light were used to create masks， and the corresponding exposure duration under grating projection were multiplied using stripe images. After a linear transformation， the obtained images were superimposed and then subjected to a gamma transformation. Finally， the obtained stripe images were used to solve the phase and calculate the three-dimensional point cloud data of the measured object surface. Experiments show that this method can overcome the lack of point cloud data caused by the local highlight of the object to an extent. For the three tested objects with high reflection surface of flange， standard block， and U-card， more than 99.8%， 98.1%， and 99.3% point cloud data can be calculated， respectively. Consequently， a more complete three-dimensional contour detection of high reflection metal surface can be realized.