This PDF file contains the front matter associated with SPIE Proceedings Volume 13069, including the Title Page, Copyright information, Table of Contents, and Conference Committee information.

Learning-based computer-generated holography (CGH) has great potential for real-time, multi-depth holographic displays. However, most existing algorithms only use the amplitude of the target image as a dataset to simplify the algorithmic process. This does not adequately consider the incorporation of angular spectrum (ASM) method into neural networks that can compute multiplanar attributes. Here, we propose a multi-depth diffraction model-driven neural network (MD-Holo). MD-Holo utilizes the weights of the pre-trained ResNet34 as initialization in the encoder stage of the complex amplitude generating network to extract more general features. Motion blur, Gaussian filtering, lens blur and low-pass filtered images are added to accommodate a wider range of images. Compared to the super-resolution DIV2K dataset alone, the use of the enhanced dataset allows both the generation of high-fidelity super-resolution images and the generalization of a wider variety of images. Simulations and optical experiments show that MD-Holo can reconstruct multi-depth images with high quality and fewer artifacts.

Light field 3D display can reconstruct 3D scenes in space through light modulation devices. All the information required for light field reconstruction is provided by an image called 3D content. However, the procedure of 3D content acquisition is complex. With increased viewpoints and higher resolution of the light field 3D display, a real-time 3D acquisition is difficult to achieve. In this paper, a real-time 3D acquisition method for light field 3D display based on the calculated depth map is proposed and a real-time 3D acquisition system is also developed. The average frame rate of the presented system is 13.66 frames per second with a resolution of 3840×2160 pixels.

As an important phase unwrapping method in Fringe Projecting Profilometry, the geometric constraint enables pixel-by-pixel computation while avoiding the projection of additional patterns. However, its measurement depth is limited to one fringe period and influenced by multiple factors from the camera-projector system. There is still a lack of a generalized model that can describe the measurement depth. In this paper, to completely understand the variables related to the measurement depth, we analyze the unwrapping process of geometric constraint and derive a generalized model. To validate the key variables in the model, we establish a virtual camera-projector system and conduct simulations. Based on the simulation results, we propose four ways to extend the measurement depth, including (a) selecting a reference plane farther away from the camera, (b) designing the fringe patterns with fewer periods, (c) using the shorter focal lens and (d) reducing the camera-projector distance.

Simultaneous Localization and Mapping (SLAM) has been widely used for indoor building information modeling (BIM) applications. However, the data perceived by traditional sensors is not satisfactory, which also limits the capabilities of SLAM. Recently, high-quality fringe projection profilometry (FPP) sensor has been introduced into SLAM for indoor applications. In this paper, we first introduce the workflow of FPP-SLAM and then verify FPP-SLAM on BIM of cultural heritages.

Fringe projection profilometry (FPP) is one of the most widely used optical three-dimensional (3D) perceiving techniques. However, in the realm of indoor 3D perceiving, achieving high-resolution data proves challenging due to the inherent trade-off between sampling resolution and measurement scale. This paper introduces an adaptive-resolution-based method to address this challenge. Specifically, the approach leverages the super-resolution reconstruction technique to enhance the resolution of captured fringe patterns, where an end-to-end fringe pattern super-resolution network (FPSRNet) is constructed to adaptively achieve different super-resolution values. The desired high-resolution 3D results can be reconstructed from these new patterns. Experimental results verify the effectiveness of the proposed method in indoor 3D perceiving.

In this paper, a multi-view structured light three dimensional (3D) imaging system is descried, which can perform rapid 3D reconstruction of the human body and automatically extract anthropometric data from it. This system contains 12 sets of 3D imaging sensors distributed on four pillars. Each 3D imaging sensor consists of a binocular stereo system, an Infrared laser projector and a synchronous control system based on the Field Programmable Gate Array (FPGA). The projector provides phase-shifting fringe patterns and gray code for the binocular stereo system to make 3D reconstruction. The FPGA control system enables the sensor to achieve high speed scanning. A two-step calibration method is used to calibrate the internal and external parameters of each 3D imaging sensor and external parameters between these sensors. After the 3D human body data acquisition, major body joints will be extracted as key-points. In this process, the initial location of these key-points are extracted based on a deep learning method, and then they are further corrected with local point cloud analysis. With the assistance of these key-points, the anthropometric data, such as distances (lengths, breadths, heights) and circumferences of the human body, can be calculated from its 3D data. Based on the techniques described above, the multi-view 3D imaging system can complete the whole body scanning in 2 seconds and automatically measure more than sixty dimensional data after analyzing the reconstructed 3D human data.

In order to solve the problem of high-precision registration of point cloud pairs with known correspondence under constrained conditions, a point cloud pairs registration method based on directed distance function is proposed. The point-to-point directed distance function is constructed by first order linear Taylor expansion, and LM algorithm is used to solve the least squares problem. An improved particle swarm optimization (PSO) algorithm is introduced to find the optimal normal vector between point pairs, so that the directed distance function iteratively approximates the sum of the minimum Euclidean distances between point pairs. The experimental results show that compared with SVD method, the proposed method is close to its accuracy under short distance variation, while the calculation accuracy is obviously improved under large distance complex variation.

3D reconstruction based on infrared images can restore the 3D model with temperature information, which helps observe the appearance and structure of the target more comprehensively and has potential application in the fields that need temperature detection, such as environmental monitoring and risk detection. This paper studies the 3D reconstruction of the infrared scene based on a UAV platform. An infrared camera captures the images, and the incremental SFM algorithm is used for further 3D reconstruction. The experimental results show that the 3D point cloud from the infrared image is satisfactory, the 3D reconstruction of terrain such as roads, houses, and trees is successful, and the 3D heat distribution of the scene can be displayed.

In near-field photometric stereo vision, the accurate calibration of the light source position directly affects the precision of the reconstruction results. Traditional calibration techniques rely on a highly reflective sphere and exploit the specular reflection properties at highlight points for determining light source positions. However, in real-world scenarios, nonuniform lighting conditions often leads to errors in extracting the sphere's image edges, affecting the accuracy of light source position calibration. Therefore, we propose a method for calibrating light source position based on a novel target. This method involves detecting planar reference points that are less sensitive to the lighting conditions and leveraging the pose relationship provided by the planar reference points to effectively overcome the adverse effects of lighting conditions on light source position calibration. Experimental results demonstrate that this method significantly enhances the precision of light source position calibration under non-uniform lighting conditions, eliminating the constraint of light source position calibration on specific lighting environments.

In situations where the object being captured or the camera itself is in motion, object tracking using conventional attached markers or detailed textures is useful to keep capture the object with high accuracy; however, tracking accuracy has been degraded in situations where texture features are scarce or where it is difficult to attach markers. Therefore, in this study, the temperature rise corresponding to the laser irradiation time to the object is detected by a thermal camera, and the points where the temperature reaches or exceeds a threshold value after the laser irradiation ends are used as markers. After the temperature drops, laser irradiation is performed again to repeatedly generate markers. Furthermore, by controlling the irradiation point with a 2-axis galvanometer mirror, a marker that can code arbitrary shapes and information can be generated on a 2-dimensional plane. In our experiments, we irradiated a red semiconductor laser onto black paper, and found that an irradiation time of 10 ms was the most efficient in terms of heat dissipation time, enabling the simultaneous generation of up to 33 markers. As a result, it was found that it is possible to draw distinguishable characters and graphics such as lines and circles by continuous drawing. Future plans include application to non-paper based on additional physical property investigations and actual tracking applications.

Autofocus plays an important role in microscopic imaging. As an extension of image-based methods, learning-based methods make real-time autofocus possible. The recently proposed learning-based autofocus methods achieved promising results in estimating defocus distance. However, the focusing accuracy depends partly on the feature extraction ability of the network model, and what features are specifically extracted by the network contributed to its success remains a mystery. In this paper, a single-shot microscopic autofocus method was proposed, which predicts the defocus distance from a single natural image, to improve the model's ability to extract image detail features. Furthermore, we validate that the neural network model mainly predicts the defocus distance by focusing on the sharpness of texture and edge features, and visualize the weight of the predicting results. A realistic dataset of sufficient size was made to train all models. The experiment shows the proposed network model has better focusing accuracy compared with other models, with a mean focusing error of 0.44μm, and pays more attention to the texture and edge features.

Fringe projection profilometry has been applied to measure 3D information of fingertip and collect contactless 3D fingerprints. When low-resolution (LR) camera is used in the system due to reasons such as cost, the captured fringe patterns may appear blurry, which results in less obvious contrast between valleys and ridges in the reconstructed contactless 3D fingerprints. To address this issue, we introduce an unsupervised super-resolution (SR) method that solely relies on low-resolution fringe patterns. Our approach combines a two-loop generative adversarial network. In the forward loop, a binarized interpolation loss function is designed to ensure that the upsampling generator preserves ridge and valley details. In the backward loop, the discriminator ensures that the fringe patterns produced by the downsampling generator are both repeatable and similar to the original fringe patterns. Finally, the fringe patterns are reconstructed to obtain 3D fingerprints. Experimental results demonstrate the advantages of our proposed method.

Existing three-dimensional scanning techniques enable the acquisition of dense point clouds representing the surface of the scanned object. However, the voluminous nature of unordered point cloud data leads to extended data processing times, necessitating the utilization of specific data structures for the management of large-scale point clouds. Addressing the performance degradation issue of prevalent point cloud data structures when dealing with a large quantity of points, this paper initiates a comparative analysis of common point cloud data structures, encompassing grid-based, quadtree, and k-d tree (k-dimensional tree) indexing methods. Through theoretical derivations, an examination of the time complexities of various data structures is undertaken. Building on this theoretical foundation, an empirical quantitative assessment of the real-world performance of distinct data structures is executed. Leveraging the insights gained from these analyses, this paper further capitalizes on the inherent shape characteristics of empirically acquired point cloud data to introduce a novel three-tier hybrid indexed point cloud data structure, accompanied by its corresponding algorithmic functionalities. This innovative structure amalgamates grid-based, quadtree, and k-d tree indexing strategies. Empirical findings demonstrate that, when applied to large-scale point clouds, the proposed three-tier hybrid indexed data structure exhibits enhanced indexing establishment speed and neighborhood search velocity compared to conventional algorithms. Thus, this work establishes a foundational data structure support for subsequent processing and application of large-scale point cloud data.

In the manufacturing industry, the high-precision and high-efficiency dimension measurement of the large componentsisan important guarantee to improve product quality and production efficiency, but the traditional contact measurement method is low efficiency, poor accuracy, time consuming and vulnerable to human factors interference, has beenunableto meet the requirements of rapid and accurate measurement. To solve this problem, based on the three-dimensional point cloud data of large components, this paper studies the geometric feature extraction and dimension measurement methods of components. The 3D point clouds of components are preprocessed by establishing topological relationship, estimating surface normal vector and point clouds filtering for noise reduction. Geometric features of preprocessedpoint clouds are extracted, including point clouds with straight line features such as side edges and point clouds with circulararc features. The specific steps include extracting key planes by RANSAC, extracting edges of planes based onnormal vector estimation, retaining point clouds with geometric features, and dividing point clouds by Euclidean clustering. After that, the extracted point clouds with geometric features are synthesized into straight lines or circles to measurestraightness and roundness. Besides, a method is proposed to search adjacent points on the linear point clouds in order tomeasure arc length and analyze error sources and accuracy. The experimental results show that the measurement methodproposed in this paper can achieve high precision dimension measurement of the components.

Foreign object detection and localization is of critical importance in various real-world scenarios. For example, a foreign object in an automatic assembly line could result in severe dangers. In this paper, we propose a two-stage deep learning method to identify and localize foreign objects in a working environment. It includes two major stages, i.e., detection and localization. In the detection stage, an advanced anomaly detection model is first used to identify unknown object candidates. However, the unknown objects detected in this stage might include potential normal classes. Subsequently, we use the working environment data to train a YOLO model to filter out the false positives in the potential unknown objects. In the localization stage, we use K-Means++ to cluster a heatmap generated in the first stage and extract activation points with highest activation scores which are fed into an advanced segmentation model for accurate segmentation and localization of foreign objects. We have conducted experiments to validate the performance of the method in an experimental setup. The developed mode can well adapt to various scenarios in manufacturing automation.

In incoherent digital holography (IDH) and in any imaging technique, the lateral and axial resolutions are intertwined and consequently changing one characteristic affects the other. In this study, we present two new hybridization techniques HM-1, and HM-2 for IDH, one for real-time and another for post-recording of holograms respectively, to engineer the axial resolution independent of lateral resolution. Two optical functions namely a lens and an axicon, with a low focal depth and a high focal depth respectively are considered. In both hybridization techniques, the axial resolution can be tuned between the limits of the axial resolutions of lens and axicon, while maintaining a constant lateral resolution. In HM-1, the axial resolution was engineered using a special phase mask designed using a modified version of Gerchberg-Saxton algorithm that can generate a spherical beam and Bessel beam for every object point and create self-interference between them. By controlling the strengths of the two beams, the axial resolution can be tuned without changing the lateral resolution. This method requires an active optical device such as a spatial light modulator. HM-2 involves two recordings of the same scene, one with a lens and another with an axicon which are then combined after recording. By controlling the weights of the two recordings, the axial resolution can be tuned between the limits of lens and the axicon independent of lateral resolution. In this case, passive diffractive or refractive optical elements are sufficient. Both hybridization techniques are implemented in indirect imaging mode consisting of three steps: recording point spread hologram, object hologram and reconstruction by Lucy-Richardson-Rosen algorithm.

Recently, the characterization of marine objects, populations and biophysical interactions have become crucial within the research community. In this study, we leverage digital holographic imaging systems and deep learning networks to classify three distinct types of micro-algae: Chlamydomonas, Scenedesmus armatus, and Scenedesmus_sp-L. We employed reconstructed digital holographic images and deep learning to identify the results from both approaches. The integration of holographic imaging holds promises in replacing expensive characterization systems like AFM, x-ray diffraction, and Raman spectroscopy, offering a more costeffective solution. In our system, we utilize in-line microscopic digital holographic imaging to record and reconstruct images of the algae specimens. An essential advantage of holographic techniques is that they do not require intact samples of the specimens for effective object identification. To further enhance the process, we combined deep learning algorithms with holographic imaging, capitalizing on the advanced computers. This combination enables highly effective characterizing and classification of different types of algae. These innovative approaches pave the way for exciting advancement in marine research and monitoring.

A multiple-image cryptographic authentication scheme based on phase-only holograms is proposed in this paper. First, computer-generated holograms of original images to be authenticated are encoded to phase-only holograms using the Floyd-Steinberg error diffusion algorithm. Second, each phase-only hologram image is randomly sampled as the sparse representation with the help of its random binary mask. Finally, the phase-only ciphertext containing the information of original images is obtained by integrating all sparse representations using their binary masks. The existence of each original image can be verified by calculating the nonlinear correlation map between it and its corresponding decrypted result. High security level of this cryptosystem can be achieved by considering random binary masks as secret keys. This work provides an effective alternative for the related research based on computer-generated holograms.

Traditional numerical reconstruction methods in digital holography are faced with problems such as inaccurate and time-consuming unwrapping or the need to capture multiple holograms with different diffraction distances. In recent years, deep learning, as a new and effective optimization tool, has been widely used in digital holography. However, most supervised deep learning methods require large-scale paired data, and their preparation is time-consuming and laborious. Here, we propose a new deep learning approach that can use less unpaired data to train neural networks, thereby reducing the need for labeled data. This method can reconstruct complex amplitudes for holographic reconstruction and generate synthetic holograms at the same time. The reconstructed complex amplitudes have higher image quality, while the generated holograms can reconstruct the complex amplitudes successfully

Defect detection is crucial to the manufacture and evaluation of materials. However, it is still a great challenge to detect the defects in a wide field. In this paper, the two-dimensional (2D) digital multiplication moiré method is presented. The point defects of the crystal are detected visually by employing digital image processing. We mainly discuss the applications of this method to detect the defect and measure the strain in the silicon (Si) single crystals. The strain distributions in the main directions of Si single crystals are measured, and the point defects are detected. Point defects are easier to observe when the atomic structure is amplified using 2D multiplication moiré. The 2D multiplication moiré method that has been used for the point defects detection in Si single crystals described in this paper also lays an important foundation for the detection of strains and defects in the crystal structure of other materials.

The multiplication sampling moiré (MSM) method achieves a strong noise-immunity deformation measurement by performing phase analysis of the second harmonic of grating patterns, which surpasses the limitation of the conventional sampling moiré method that produces phase errors when the first harmonic is submerged by the background noise. In this study, the multiplication sampling moiré method was utilized to investigate the fracture behavior of a [±15°]2s carbon fiber reinforced plastic (CFRP) laminate specimen under different tensile loads. The full-field microscopic strain distribution maps, including the normal, shear, and principal strains, were successfully measured on the cross-section of the CFRP laminates with fiber discontinuities. The results show strain distribution characteristics before and after transverse crack occurrence in the matrix resin region of the CFPR laminates, and the changes in shear strain at the interlayer interfaces before and after the emergence of delamination. The MSM method holds promise for evaluating mechanical properties, fracture behavior, characterizing strain distributions, and residual stresses in deformation measurements of various structural and composite materials.

The damage behavior of carbon fiber reinforced plastics (CFRP) is highly complex due to the overlapping of various damage forms. To elucidate this deformation behavior, it is crucial to evaluate the micro-scale deformation distribution before the occurrence of each type of damage. The sampling Moiré method has been recognized as an effective experimental technique for analyzing such deformation behavior. In this study, we used the sampling Moiré method to calculate the micro-displacement distributions of [±45°]_4s CFRP laminates during a three-point bending test under microscopic observation. Additionally, a finite element model was created under the same conditions to compute the displacement distribution using the finite element method (FEM). A comparison of the displacement distributions obtained from both experimental and simulation methods confirmed their consistency. The simulation results also revealed the difference in CFRP displacement distribution characteristics with and without interlayer resin, which manifested that the presence of a resin layer between the CFRP layers induced a distinctive wave displacement distribution when subjected to a three-point bending load.

Accurate deflection measurement is vital in evaluating the structural integrity of transportation infrastructures, with bridges being fascinating. In this study, we propose a novel image stabilization technique integrated into the sampling moiré method, which leads to a dependable approach for measuring bridge deflection through drone aerial photography. Our experimental verification entailed conducting drone tests on an actual bridge, utilizing a passing test vehicle, and the results showcased deflection measurements comparable to those obtained through conventional methods. This newly developed technology eliminates the need for ground-fixed cameras mounted on tripods, thus enabling precise deflection measurement at the millimeter level for bridges in challenging environments, including marine and mountainous areas.

RBCs are essential for carrying oxygen throughout the body. Maintaining human health requires an understanding of the various RBC types, their structural defects, and the difficulties in identifying these abnormalities. RBCs are commonly divided into sickle cells, regular disc-shaped erythrocytes, and other physical variations. Hemoglobinopathies including sickle cell disease, thalassemia, and genetic spherocytosis, as well as acquired syndromes like anemia, which can be brought on by dietary shortages or long-term illnesses, are only a few examples of the wide range of RBC abnormalities. Advanced imaging techniques are necessary for identifying and characterizing these anomalies. Label-free, non-invasive, and high-resolution imaging of RBCs is made possible by QPI techniques like the Transport of Intensity Equation (TIE). In our work, with the use of TIE-based 3D QPI, we have extracted quantitative features like cell volume, cell height and cell surface area of human RBCs from the captured images. This method enables characterization that is more accurate and diagnosis of diseases by providing insights into the structural modifications linked to RBC abnormalities

In this paper, we report an approach to realizing optical data transmission through dynamic scattering media using a pixel-to-plane data encoding strategy, and high-fidelity and high-security transmission can be realized. A signal is considered as a sequence of separate pixels which are sequentially encoded. We generate a series of 2D patterns as information carriers to be used in the optical transmission channel. To suppress noise, a differential protocol is designed and applied. In the designed optical system, numerous keys are generated to guarantee the security. The absorptive filters are used, and ciphertext can be obtained. Our experimental results illustrate validity of the method. Only when the keys are correct at the receiving end, the encoded data can be retrieved. It is expected that this approach can be useful to secure free-space optical data transmission through dynamic scattering media.

Ghost transmission through scattering media remains an open question. Here, we present high-fidelity ghost transmission through complex scattering media using random patterns as information carriers. Pixel values of an analog signal to be transmitted are sequentially encoded into random patterns by employing the untrained neural network (UNN). In optical experiments, the laser beam illuminates the designed random patterns and passes through complex scattering media. The light intensities recorded by a single-pixel detector are used to retrieve the encoded analog signal. Experimental results demonstrate that the developed pixel-to-plane pattern encoding can achieve high-quality ghost transmission through complex scattering media. The method provides a solution for ghost transmission in complex environments by applying UNN for optical encoding.

High-fidelity information retrieval through complex scattering media has been a challenge. To address this issue, we present a modified Gerchberg-Saxton (GS) algorithm that generates random amplitude-only patterns to serve as information carriers. The modified GS algorithm imposes a support constraint to a random pattern in the image plane and scales the amplitude of its Fourier spectrum to control the sum of the pattern. Therefore, random patterns generated by the modified GS algorithm are encoded with pixel values of the transmitted data. The patterns are sequentially displayed by a spatial light modulator, and optical wave is recorded by a single-pixel detector. Optical experiments have been conducted to evaluate the method under various conditions, e.g., dynamic and turbid water and non-line-of-sight (NLOS). It is experimentally verified that the method can realize high-fidelity ghost transmission in complex scattering media.

An optical image encryption method is proposed based on Fourier single-pixel imaging and iterated phase retrieval algorithm. First, a binary barcode image containing two groups of horizontal strips is randomly generated and used as the target image in the single-pixel imaging process. Second, considering the barcode image as the amplitude constraint, the original plaintext image is encoded to two phase-only masks using a designed phase retrieval algorithm in Fresnel domain, and these masks are applied as secret keys. Finally, the barcode image is encrypted to a series of measurements using two-step phase shift Fourier single-pixel imaging as the ciphertext. Differing other methods, the original image is not directly imaged and encoded into the ciphertext. Even the ciphertext is obtained by a cracker, only the barcode image may be further discovered. Due to two phase-only masks as secret keys, the security level of the cryptosystem can be enhanced greatly. The simulated results verify the feasibility of the proposed method, and this work provides an effective alternative for the related research.

Realizing high-quality object reconstruction in complex and dynamic scattering environments is a challenge, especially in highly dynamic scattering environments. Scaling factors can be considered to be constant in a static environment. A highly dynamic scattering environment can result in the changed scaling factors. In this paper, we report a high-quality object reconstruction method using Hadamard-based single-pixel measurement in highly dynamic scattering environments. In the proposed method, the sequence of Hadamard patterns is randomly ordered, and then the Hadamard patterns are applied to illuminate an object. The wave passes through a dynamic and turbid water tank. In this case, randomly changed scaling factors can be obtained. A temporal correction method is applied to eliminate the randomly changed scaling factors in single-pixel intensity measurements. After the temporal correction, high-quality object reconstruction is realized in highly dynamic scattering environments. Experimental results are presented to demonstrate validity of the proposed method.

The speckle effect caused by coherent light scattering from rough surfaces modulates the intensity and phase of the interference signal, which results in signal dropouts and phase noise for Laser-Doppler vibrometers (LDV), limiting the accuracy and resolution of LDV. Signal diversity has been demonstrated to significantly suppress the speckle noise in single-point LDV vibration measurements since two or more statistically independent reception channels significantly decrease the probability of speckle noise occurring. In this paper, an orthogonal interferometer based on polarization diversity is presented. We discuss the potential of suppressing the speckle noise with dynamic ellipse fitting and dual-channel data fusion. The displacement measurement of LDV is obtained by phase synchronization and a simple weighted sum method. The conclusion of this article is that signal diversity combined with a dynamic ellipse fitting algorithm is feasible to suppress speckle noise resulting from the periodic movement of the target.

Hyperspectral anomaly detection (HAD) does not require a priori information, and accurate discrimination is made by analyzing the difference between the anomalies and the background pixels. However, the bands of hyperspectral images are highly correlated with each other. There is a lot of redundant information between them, which causes the band selection to be difficult to accurately distinguish between background and anomalies. This paper introduces background purification and feature extraction strategies to increase the distinction between anomalies and background pixels. To be specific, the domain transformation extracts discriminative sample features. The row-constrained low-rank sparse matrix decomposition is utilised to obtain low-rank background matrices to construct purer background to highlight the anomalies. The sliding window strategy is adopted to divide the subspace to reduce the spatial correlation. Highly representative and low redundancy bands are selected for band selection in the local region. Finally, the local region is detected by RX and the map is obtained by domain-valued normalisation of the local results. Experiments on several HSI data sets show that the proposed method can suppress the background well. It can also make full use of the spectral information and achieves acceptable detection accuracy.

The principle of laser triangulation has been widely used in various high-precision measurement due to its advantages of high accuracy, fast response speed and simple structure. However, the laser triangulation method still has some limitations for the measurement of dynamic object, and the accuracy of laser location will be significantly reduced especially when the working distance becomes longer. We have carried out the development and performance optimization of a long-distance laser triangulation displacement sensor for the application of dynamic object. The measurement uncertainty is analyzed when the laser dithering, laser drifting and object motion are all taken into account. Then, reflectors are added to the optical path structure of traditional laser triangulation system with the aim to improving its sensitivity and expanding its working range. The experiment results show that the repeatability of our designed optical system is 2.9μm and the imaging uniformity is 86.23% when the working distance is of 1300mm.

Current requirements resulting from high quality standards and mass production call for an efficient and reliable manufacturing technology. With the aim of meeting the needs of accurate and efficient inspection of massive industrial parts, we propose a novel large-field measurement technique to measure weakly textured, complex reflective surface on large-size parts. We introduce a small-field-of-view, high-spatial-resolution binocular fringe projection local measurement system, which realizes high-precision point cloud data acquisition of industrial parts' surface based on the principle of heterodyne multi-frequency phase shift and high dynamic measurement. A new point cloud registration method based on phase matching and global marker points is proposed. A global measurement system with large field of view and low spatial resolution system is applied to track the local measurement system and realize the local area point cloud registration of large industrial components. The global marker points on the surface of the turntable jointly calibrated by the laser tracker and the vision system is presented to realize the registration of all areas of the component. A global optimization with trimmed ICP is applied for adaptive fine registration of multi-view point clouds with different overlap rates to eliminate accumulated errors. The proposed system can achieve a global measurement range of 4m×4m×2.5m, and the experimental results demonstrate the effectiveness of the proposed method for high-precision measurement of industrial complex parts with large field of view, which effectively avoid the transmission error and benefits the inspection and manufacturing of industrial parts.

To address the challenges in evaluating assembly quality for revolved thin-walled parts, a novel algorithm was developed that combines virtual assembly and quality assessment. Leveraging 3D point cloud data, the algorithm employs various techniques for accurate assembly assessment. Initial registration using an Oriented Bounding Box (OBB) achieved rough alignment, followed by precise registration using point-to-plane Iterative Closest Point (ICP) to minimize RMS error to 0.173mm. Contour points on cross sections were extracted through resampling. Transverse section contours were effectively fitted using least square circle fitting. However, due to its randomness, least square ellipse fitting for longitudinal section contours struggled to meet precision standards. To quantify part manufacturing errors, RMS radial distances from longitudinal section contour points to theoretical ellipses were utilized. The algorithm then calculated the assembly sequence with the least error. Detection of assembly interference was realized by measuring distances between transverse section contour boundary points. Simulated point cloud data validated the algorithm's efficacy in effectively assessing part quality, optimizing assembly sequences, and identifying assembly interference. By combining innovative registration, contour analysis, and error quantification techniques, the algorithm offers a promising solution for ensuring assembly quality and reliability, addressing the challenges posed by irregular revolved thin-walled parts.

Laser Induced Breakdown Spectroscopy (LIBS) is widely used in analytical chemistry, biomedicine, and environmental applications due to its real-time detection of multi-elements. However, the main challenge for LIBS application lies in its low detection sensitivity, especially for liquid sample analysis. When plasma is generated by a nanosecond laser pulse inside the liquid sample, fast quenching of the plasma occurs, and atomic emission intensity becomes weak with a short lifetime. Furthermore, the creation of surface fluctuations during laser ablation reduces the reproducibility of the signal. Researchers started exploring the possibility of improving the plasma signal by investigating different sampling approaches for liquids to increase the signal-to-background ratio of the signal. Liquid LIBS has the potential to become one of the best ultra-sensitive elemental characterization methods by standardizing the technique and making it applicable for potential industrial applications. In this context, this paper investigates two different sampling approaches for liquid LIBS signal enhancement. The experimental configurations, optimization of the experimental parameters, and the limit of detection of the sampling approaches of the proposed LIBS are detailed, followed by its potential application in vertical hydroponic farming.

Internal surface condition of mould will affect the metalworking product quality after the casting process. Traditional visual inspection devices cannot keep working when the temperature is beyond 70°C. A vision inspection system protected by a cooling housing was introduced in this paper. The system was designed to provide quality images for defect inspection needs while enduring high temperatures. Illumination strategy consisting of lighting position and orientation was developed to cover 360° inner surface and accentuate defects in images. The housing protected vision system would be motorized into the mould to user-specified position and capture images with trigger mode. Developed algorithm would process images and output inspection results in the backend. Overall inspection of 2-meter mould could be completed within 100 seconds. The developed vision system is fast and low-cost, which can be easily implemented on site and provide defect inspection for cast mould with high temperature.

Zernike polynomials are a complete set of continuous functions orthogonal on the unit circle, commonly used for wavefront fitting and analyzing wavefront properties. Zernike polynomials have the special properties of orthogonality and normalization within the unit circle, which makes them widely used in wavefront fitting and reconstruction. In addition to circular pupils and circular elements, non-circular shapes such as squares ellipses are usually found in optical systems. For non-circular wavefronts the Zernike polynomials lose their orthogonality, which also leads to coefficient coupling thus affecting the effectiveness of aberration removal. This paper presents the method based on the Gram–Schmidt orthogonalization technique to orthogonalize Zernike circular polynomials over the non-circular region through a series of matrix transformations. The proposed method can obtain Zernike wavefront fitting results for arbitrary shape wavefront without deriving the corresponding set of polynomials. Separate wavefront fits were conducted utilizing various wavefront shapes, and the results were analyzed. The fitting of non-circular wavefronts is realized in experiment using orthogonal Zernike matrix, which verifies the effectiveness of the proposed method.

Defect detection is an important part of heritage conservation and speckle pattern interferometry is a common technique for inspecting surface and internal defects. A speckle pattern interferometry coupled with pulsed laser system for non-destructive detection and prediction of different size cracks is introduced and tested. According to photoacoustic effect, ultrasonic waves are generated by pumping a pulsed laser beam to the rear surface of the sample. The ultrasonic waves serve as carrier propagating from the rear surface to the front, thus conveying deformation and crack size information. For obtaining information about simulated crack size, the front surface is detected by a speckle pattern interferometry system. And simultaneously, the generated ultrasound waves are detected by air-coupled transducer. In this study, the introduced system and method were validated by detecting medium density fiberboards with simulated cracks of different width and depth. Differentiated speckle pattern and ultrasound signals are compared and combined to indicate crack presence and to figure out different crack size qualitatively.

The visual measurement techniques play an important role in underwater three-dimensional (3D) topography reconstruction. Line structured light measurement are widely used in visual 3D information acquisition. Aiming at the refractive nonlinear imaging effects and image degradation problems in underwater measurements, this paper proposes a physical enhancement-based underwater nonlinear refraction measurement model (peUNR). First, the underwater images are corrected by physical enhancement based on scattering imaging theory to improve the image recognition. Second, a refraction relationship for underwater visual measurements is developed by ray tracing. In this model, the light plane is unchanged before and after reaching the surface of the underwater target. As a result, the system parameters can be calibrated in air, allowing high-precision 3D reconstruction underwater without the need for underwater calibration. At last, in order to reduce the dependence of structured light stripe extraction on image quality, a multiline rotating structured light projection mode is designed by using binary coding, which fulfils the need for fast and portable measurements and improves the measurement efficiency and resolution simultaneously. The experimental results of underwater measurement indicate that the residual error of the fitted plane obtained by the peUNR model is 1.81 mm in turbid water with a sand content of 125 g/m³, which satisfy the requirements of underwater 3D high-precision measurement.

In certain scenarios, due to the limitation of a single camera in capturing the entire scene, we need to employ multiple cameras to gather information and seamlessly fuse multiple local images into a panoramic image, showcasing significant potential across a wide range of application domains. This paper focuses on the theory and practice of image stitching in fixed wide-field-of-view scenarios. Firstly, we delve into the fundamental principles and methods of image stitching, including feature extraction, feature matching, and image blending. Secondly, we provide an overview and analysis of current state-of-the-art image stitching algorithms, exploring their advantages and limitations, particularly in the context of fixed field of view. To address the common challenges in fixed wide-field-of-view image stitching, this paper proposes a novel stitching algorithm based on grid distortion. By utilizing multiple sets of scene images to guide the new image stitching process, we achieve higher precision in image matching and blending while retaining fine details. Experimental results demonstrate the effectiveness of our approach in a self-constructed scene.

This paper presents a method of extracting Zernike coefficients and then retrieving the phase based on deep learning. In the normal interferometry, the phase extraction algorithm is utilized to calculate the wrapped phase according to the registering phase shifting interferograms, and then unwrap it to obtain the phase map. Further, Zernike polynomials are used to fit the phase. The method proposed in this paper uses neural networks to extract 35 terms of Zernike coefficients from two random phase-shifting interferograms, then retrieve the phase based on Zernike polynomials. The method only needs two random phase-shifting interferograms, does not require phase extraction and unwrapping processing and greatly simplifies the computation for phase reconstruction. The paper presents the training processing, and provides the experimental results. The results show that the proposed method can reach high precision and are more suitable for the quick testing for workshop environment.

Phase-shifting interferometry is a high-precision and commonly used phase retrieval method. In practical applications, phase shift errors are usually introduced due to factors such as environmental disturbances and phase shifter error. In this paper, we propose a deep learning method for estimating phase shift error from phase-shifting interferograms. This method manages to process the three interferograms with phase shift π/2, and uses neural network to extract phase shift errors from three interferograms. The analysis shows that the proposed method can effectively estimate the phase shift error under the noisy interferograms. This method can be used to correct phase-shift errors for phase retrieval (e.g., the least squares phase retrieval method) and calibrate phase shifters.

Modern third-generation synchrotron radiation sources provide more collimated, brighter, and coherent X-ray beams for experimental techniques. X-ray optics are the bridge between the light sources and the experimental stations. Any defect (either from mirrors or crystals) will bottleneck preventing the exploitation of the full characteristics of the source. In addition to high-quality X-ray optics, mirror mounting, and handling of thermal deformation are also of critical importance. Advanced metrology to properly exploit all the new potential of these optics is needed. Shanghai Synchrotron Radiation Facility (SSRF) has metrology labs equipped with visible-light-based measuring instruments and X-ray test beamline for in-situ metrology. In this article, we will present the current state of the art of mirrors, crystals, and diagnostics at SSRF.

The supply chain of agricultural products is intricately linked to the daily lives of people. In light of rising import and export quantities, the need for a prompt and efficient inspection system has become increasingly pressing. Without opening baskets and manually sorting, a smart inspection scheme is designed in this work leveraging X-ray images and transformer neural network. Due to its penetrating capabilities, X-ray enables a direct examination of agricultural products within a basket, a task that normal vision devices are unable to accomplish. Taking into account the varying shapes and combinations of agricultural products, we introduce a transformer-based deep neural network for type identification. Additionally, a dataset augmentation process is developed inspired by computed tomography generating 1,6000 X-ray images. Through experiments, the proposed smart inspection scheme is proven to be feasible and works efficiently. The inspection accuracy for both single-type and mixed-type agricultural products on the established dataset exceeds 90%.

In order to achieve the effective splitting of X-ray, the perfect crystal whose the space of lattice plane comparable to the X-ray wavelength can be used as beam splitter. This beam splitter utilizes the diffraction effect of Laue crystal to accurately manipulate X-ray beam. A stress-free crystal with thin thickness is crucial for high-quality X-ray splitting. The working area of crystal was thinned by acid etching. Additionally, the base of crystal was cut from a floating-zone silicon single-crystal ingot, which prevented the spread of stress to the working area of crystal in fabrication and experiment. The experiment of Laue diffraction was conducted at the synchrotron radiation facility. In order to obtain the inherent rocking curve and consistent imaging field of view, a non-dispersion configuration was employed to match the energy bandwidth of the incident beam with the Laue diffraction crystal. Then diffraction splitting within the energy bandwidth of the Laue crystal was achieved by utilizing a collimator to reduce the divergence of the incident beam. The fine structure of the diffraction curve was measured experimentally, and the slope error of linear fitting between the high-angle and low-angle positions of peaks is less than 0.4%, which satisfies the requirement of stress-free diffraction splitting. The design and characterization of this Laue diffraction crystal provide technical support for various applications, such as X-ray ghost imaging, X-ray multi-projection imaging, and beamline measurement at wavelength.

Continuous Inspection and maintenance of high performing modern structural composite structures are essential to ensure the safety and efficiency of the any industry. Unlike conventional metals, the damage of fibre-reinforced polymer matrix composite materials that are commonly used in structural components is relatively difficult to detect, given the various micro-constituents present. Especially, carbon fiber-reinforced polymer and glass fiber-reinforced polymer laminate can produce no-visible surface damage while sustaining internal delamination and fiber failures upon experiencing Low-Velocity Impact (LVI) forces. Barely Visible Impact Damage (BVID) is one of important damages that is tedious to detect with non-destructive methods as the damage location and intensity is unknown to the operator unless detected using Non-Destructive Techniques (NDT). Therefore, Structural Health Monitoring (SHM) is an active system that provides constant surveillance of the component’s vitals in operating conditions, thereby reducing the structural Meant Time To Repair (MTTR). In this work, piezoelectric based sensors are embedded into a composite laminate with electrical cables for voltage detection under LVI impacts. Experiments were conducted with an array of sensors at various locations. The measured signals are analyzed for their amplitude with reference to the embedded location to determine the damage intensity and impact location. A Machine Learning (ML) model is developed to provide a predictive method for SHM of the composite structure for impact damage. Besides, mechanical tests are also conducted to prove the compatibility of the embedded sensors in the host structure in order to check for the knock down in the safety factor of the component due to the presence of a foreign object in the material system. The result from this study aims to develop a solution for a structural smart skin to increase the safety & reliability of composite components, assist repair technicians in reducing the time take to detect the damage location and the degree of repair required structural, as well as to enable the predictive maintenance tool for an efficient and environmentally conscious industry that reduces the material consumption and wastage during the repair stages.

Video Motion Magnification (VMM) has gained considerable attention in the field of engineering measurement due to its impressive ability to amplify subtle motions. However, traditional algorithms often suffer from image artifacts and noise due to improper parameter settings, especially when dealing with weak motion. This paper introduces an improved VMM method that addresses this limitation by incorporating the Digital Image Correlation (DIC) technique. The proposed method utilizes DIC-measured image displacement results to analyze the dominant motion frequencies. Based on this analysis, the parameters of the VMM time domain filter are set accordingly. This approach enables motion magnification in videos while preserving image details and reducing noticeable artifacts. Simulation experiments conducted on an indoor precision displacement platform demonstrate the effectiveness of the proposed method. It eliminates the need for repetitive manual parameter adjustment through trial, producing clear and amplified motion videos. Additionally, it enables accurate measurement of small-scale motions. Overall, the proposed method enhances the performance of VMM by leveraging DIC and provides a more reliable and efficient approach for motion magnification in engineering measurement applications.

Coherent Extreme Ultraviolet (EUV) and x-ray sources have many applications in industry and fundamental research. Although these sources are available in the large facility centres such as synchrotrons and free electron lasers, high-harmonic generation (HHG) can provide us a tabletop approach. However, coherent EUV and X-ray sources based on HHG normally suffer a very low photon flux. Quasi-phase-matching (QPM) techniques can be applied to enhance the HHG photon flux. Here, we apply multimode quasi-phase-matching for HHG in gas-filled hollow core waveguide in the simulation. The quasi-phase-matching is achieved by inducing mode beating between two waveguide modes. A pair of pulses with adjustable relative time delay is coupled into the hollow core waveguide. The mode beating happens exactly at the end of the hollow core waveguide by controlling the intermodal delay. The simulation shows that the photon flux of HHG is much enhanced by intermodal delay controlled QPM. These simulation results are very important for developing the HHG based coherent EUV and x-ray sources with higher photon flux.

The qualitative and quantitative assessment of amino acids holds immense significance in the realm of life sciences. It assumes a pivotal role in comprehending intricate biological processes and, importantly, in the early detection of potential disease developments, as amino acids serve as precursors for numerous biomarkers. Our amino acid profiling method offers remarkable versatility in this context, positioning it to scrutinize amino acid profiles in body fluids. Our study introduces an innovative approach to simultaneously detect amino acids through high-performance liquid chromatography combined with laser-stimulated fluorescence, displaying a picomolar detection limit. By harnessing the power of laser-stimulated fluorescence detection, our approach offers a novel pathway to enhance the understanding of amino acids' roles in health and disease.

We present results from the studies on the generation of laser-induced periodic surface structures (LIPSS) during the interaction of femtosecond (fs) amplifier pulses (1 kHz, 800 nm) with a copper target in the form of a non-diffracting, propagation invariant, zeroth-order Bessel beam. We performed laser ablation in the air (LAA) with the focused, axicongenerated Bessel beam (~50 fs). We have investigated the novel, highly uniform, large-area, ladder-step-like nanostructures formed through laser-matter interaction. The target material used was copper (Cu) and the experimental process involved imprinting the beam profile on a single spot of the Cu surface, followed by overlapping two ablation zones (achieved through scanning the sample). Nanoscale topographical developments were extensively characterised using field emission scanning electron microscopy (FESEM). The investigation focused primarily on the central lobe ablated region and concentric rings-ablated exciting patterns. Remarkably, the raster scan ablation created ladder-like periodic surface structures with sub-15 nm spherical and aspherical nano growths. Energy dispersive X-ray (EDX) mapping confirmed the elemental distribution and surface oxidation in the engineered areas. Subsequently, the laser-engineered plasmonic surface was engaged as a surface-enhanced Raman scattering (SERS) substrate to detect traces of secondary explosives [e.g., Tetryl (TL) - 2,4,6-Trinitrophenylmethylnitramine and pentaerythritol tetranitrate (PETN)] in real-time. The obtained SERS spectra of TL and PETN unveiled signature Raman modes/peaks at 1357 cm^-1 and 1292 cm¹, with the lowest possible detectable traces of 5 µM and 1 mM concentrations, respectively. © The author(s).