The structured illumination microscopy (SIM) technique, when applied under low photon efficiency, provides an effective solution for rapid live-cell imaging, thereby enabling the investigation of dynamic cellular processes. However, noise interference during the acquisition process significantly hinders the reconstruction of SIM images, leading to substantial artifacts. To address this challenge, we propose a zero-shot learning-based SIM image denoising method (ZS-SIM). This approach relies solely on a single acquisition of noisy SIM data and achieves accurate denoising through neural network training. The original SIM image stack is downsampled and interpolated to complete the resampling process, while the traditional Wiener-SIM reconstruction method is integrated to ensure physical fidelity. We introduce a symmetric reconstruction loss and a mutual constraint SSIM loss that jointly enhance training stability and accelerate convergence, as demonstrated by our convergence analysis. ZS-SIM further achieves a favorable balance between denoising quality and computational efficiency, with low model complexity and fast inference speed, making it well-suited for practical deployment in microscopy workflows. Experimental results demonstrate that ZS-SIM efficiently and rapidly achieves artifact-free, high-fidelity denoising reconstruction, making it particularly well-suited for low-photon efficiency live-cell imaging and scenarios with limited computational resources. Furthermore, by extending the method to scanning electron microscopy (SEM) data, we validate the effectiveness of ZS-SIM for SEM data denoising, significantly enhancing the performance of downstream segmentation tasks. We anticipate that ZS-SIM will play a pivotal role in low-photon efficiency imaging, driving advancements in this field and providing crucial support for rapid validation in biomedical research, thereby overcoming the challenges posed by acquisition noise.

In surface mount technology, solder paste printing quality critically affects product reliability. Accurate segmentation of solder paste regions is essential for defect detection, quantitative analysis, and process optimization. Traditional threshold-based methods lack robustness under varying surface textures and lighting, while deep learning approaches require large annotated datasets and expensive hardware, limiting their use in cost-sensitive manufacturing. We propose a fast, annotation-free segmentation framework based on parameteric multimodal learning, integrating RGB color with 3-D height data. Height priors generate an initial mask, followed by a lookup table–based parameteric color model that adapts to different printed circuit board types and batches. A convolutional feature fusion operator then constructs a joint height–color probability space, suppressing interference from substrate variations and uneven illumination, yielding a refined probability map for final segmentation. Tests on a 3D-solder paste inspection industrial dataset achieve 96.0% mean intersection over union and 98.8% pixel accuracy, matching state-of-the-art deep learning performance while greatly improving efficiency and suitability for real-world deployment without annotated data.

Inindustrial vision systems, image degradation due to noise, illumination variation, and optical aberrations undermines the reliability of downstream tasks, such as defect detection and parameter measurement. This article proposes PhySISR, a physics-consistent self-supervised single-image super-resolution framework tailored for industrial applications. It integrates region-aware noise simulation, Airy-disk point spread function-based degradation modeling, and structure-aware loss design to restore fine details and edge structures without any labeled data. A lightweight network is further designed to reduce parameters and inference latency, ensuring deployment efficiency. To validate the method, we construct and release SMT-ImageSet, a real-world industrial dataset captured from surface-mount equipment under diverse imaging conditions. Experiments demonstrate that PhySISR outperforms representative supervised and self-supervised methods in structural recovery, edge clarity, and downstream tasks, such as binarization and parameter extraction, showing strong applicability for practical industrial image enhancement.

In printed circuit board assembly (PCBA), placement route schedule (PRS) significantly affects assembly efficiency of the beam head placement machine. The PRS is typically solved by decomposing it into placement point assignment problem (PPAP) and beam heads sequencing problem (BHSP). This article first proposes a deep reinforcement learning framework to tackle PPAP, which is a key determinant of overall process quality. Then, to mitigate the impact of placement position and angle on assembly efficiency, a dynamic programming-based beam head sequencing algorithm is introduced to solve BHSP. Since component types and placement point assignment states vary across different pick-and-place cycles, a dynamic combinatorial mask encoding method is proposed to effectively extract feature information between placement points. Inspired by the beam head placement process, a decoder that combines gated recurrent units and an attention mechanism is finally introduced, which fully utilizes historical node information to predict the next node. Experimental results demonstrate that the proposed method reduces PCBA routing distance by an average of 4.62%, outperforming other State-of-the-Art approaches.

Single image super-resolution (SISR) seeks to reconstruct high-resolution images from low-resolution inputs under the tight latency and memory budgets of edge devices. Recent visual state space models enable linear-time sequence modeling, but they often approximate two-dimensional dependencies through multiple directional scans, which increases computation and memory. We propose a Multi-View Single-Scan Visual State Space Network (MVSSN) for efficient image super-resolution. MVSSN contains three main components. A shuffled input stacking module (SISM) organizes replicated and shuffled input channels before a lightweight projection, where the learned projection filters help produce less redundant shallow responses. A multi-view single-scan block (MSSB) uses an alternating scan axis and invertible geometric transforms to change the serialization order observed by one selective scan in each block. Across stacked blocks, this provides a lightweight cross-layer approximation to multi-view context modeling. A multi-scale local feature block (MLFB) complements global aggregation with depthwise convolutions of complementary receptive fields and a compact MLP to restore local details. Experiments on standard SISR benchmarks show that MVSSN achieves competitive or better PSNR and SSIM with fewer than one million parameters and low FLOPs. Additional ablations, RealSR evaluations, and downstream detection and segmentation studies further examine its efficiency and practical behavior. We also discuss the limitation under unknown real degradation, where bicubic-trained models may still suffer from domain gaps.

Structured light imaging is a typical technique for industrial 3-D microscopic measurement. Extensive research on structured light codecs has been conducted to accurately correlate camera and projector pixels. However, these methods suffer significant degradation when measuring low-reflectivity and complex surfaces. This article introduces a deep correlation-based cascade structured light network (CasSLNet) that utilizes deep phase and column features to calculate correspondences at the subpixel scale. To mitigate the huge computational cost of full correlation, a coarse-to-fine approach is proposed. Specifically, multiscale features from the camera observation sequence and the 1-D encoding pattern are extracted through a pseudosiamese network, and cascade cost volumes are constructed. An initial column map is then regressed from the low-resolution column cost volume. Based on this, an iterative update operator is introduced to refine initial estimates, resulting in a full-resolution column map. Furthermore, a structured light dataset has been collected and experiments have been conducted on a typical structured light imaging platform. Experimental results demonstrate that CasSLNet outperforms both traditional and state-of-the-art deep learning-based methods.

Modern surface mount circuit board assemblies require more advanced defect detection methods. While deep learning algorithms have great potential for PCBA inspection, their detection accuracy in complex background situations is still limited. To overcome this problem, we propose a deep learning-based PCBA detection model (PCB-Net) to achieve accurate classification and localization of components on PCB. Firstly, for similar objects under complex background interference, this paper proposes a backbone network consisting of a generalized efficient aggregation network and a context converter to effectively extract local and global information. This integration aims to enhance the expressiveness of the network. Secondly, a multi-scale attention mechanism is designed to improve the feature extraction ability of the network on the target and suppress the interference of complex backgrounds. Finally, a C2fHB lightweight module was designed to improve the model's extraction of component features using the HorNet structure. Experimental results show that our proposed model is an effective PCBA detection method, as it can accurately detect tiny components in complex backgrounds, efficiently obtain component class and location information, and remain detection efficiency.

Vision inspection of surface mount devices is a crucial task in the operation of mounting machines. Components featuring dual-row asymmetrical pins pose a significant challenge in terms of identification due to their irregular and asymmetrical characteristics, exemplified by components like small outline transistor (SOT) and transistor outline (TO). These components exhibit variations in pin numbers, orientations, and sizes, necessitating users to provide extensive information for accurate identification using conventional methods. To address the need for fully automated recognition of dual-row asymmetric components with arbitrary orientations, a novel identification approach based on a pin search algorithm is proposed. This method comprises three main components: pin extraction, pin grouping, and subpixel refinement. Initially, connected component analysis technology is employed to extract all pins. Subsequently, a pin search algorithm is introduced, leveraging the positional and size relationships between pins to automatically detect potential pin distribution patterns. Finally, the Zernike moment method is utilized to enhance edge optimization to a subpixel level, thereby improving parameter accuracy. Experimental findings across various components demonstrate the efficacy of the proposed method in effectively resolving the identification challenges associated with dual-row asymmetric components at any orientation.

To meet the growing demands in the electronics manufacturing industry for smaller, more reliable, and flexible component packages, this research introduces a novel algorithm designed to automatically measure the geometric parameters of lead components. The algorithm can address the challenges associated with measuring these parameters in surface mount technology (SMT), overcoming the inefficiencies and inaccuracies of traditional manual measurement methods. This innovative approach combines an adaptive threshold segmentation algorithm with a multilevel grayscale range, enabling robust binary segmentation that adapts to variations in illumination. Additionally, by employing conditional filtering and state tracking, the algorithm facilitates the extraction of lead groups and individual leads, allowing for precise evaluation of component parameters through sub-pixel interpolated fitting. Extensive testing under various illumination conditions across multiple types of packaged and standard components has confirmed the algorithm's high accuracy and robustness. These results highlight its efficacy as a measurement solution suitable for industrial-grade applications.

This paper presents an improved version of the stochastic recurrent networks (STORN) for identification of non-linear state space systems with more complex model structures, including long short-term memory network (LSTM) and bidirectional gated recurrent unit (BiGRU). Both LSTM and BiGRU are recurrent neural networks with multiple state variables, which can store different information in the modeling of sequential data. Applying such a priori information into the prediction of data distribution can lead to better model performance. In this paper, several information fusion techniques are compared, and the effectiveness of the method is verified on three benchmark identification datasets. Our model outperforms the state-of-the-art baseline by 0.3 percent with 8 times fewer parameters.

Threshold selection for thresholding segmentation methods is a critical issue in vision-based detection of surface mount devices (SMDs). The thresholds in actual industrial equipments rely on manual adjustment in SMDs applications, which depends on the experience and ability of an operator. We propose a novel method based Q-Learning to automatically learn the threshold using a segmentation metric instead of manual adjustment, making the threshold adjustment process more intelligent. This framework regards 256 thresholds as state sets. A threshold list with a reasonable step is selected as actions set simulating an operator's threshold adjustment process. The metric difference of SMDs before and after threshold adjustment forms the reward/punishment of Q-Learning framework. Finally, a new strategy that determines the search direction according to previous reward is proposed to improve search efficiency. We conducted experiments on different chips and lights, and the results show that the proposed Q-Learning method performed better on search efficiency and segmentation than the traditional methods.

Identification of nonlinear dynamic systems remains challenging nowadays. Although the nonlinear autoregressive with exogenous input (NARX) model is flexible to describe complex nonlinear behaviors, it is critical to select appropriate model terms to obtain a parsimonious description of the system. In this article, a variational Bayesian (VB) approach to the estimation of NARX systems is developed. A sparsity-inducing prior is introduced for model parameters, and the sparseness can be automatically determined by the weighting factor of such prior. The Bayesian model for the identification problem is constructed, and an iterative model pruning strategy is formulated to remove redundant terms and address the structure selection problem. Instead of the single-point estimation, the model parameters with their uncertainties are jointly estimated under the VB framework. Finally, one numerical example and several benchmark datasets are adopted to illustrate that the developed algorithm can work promisingly.

Stitching images with parallax for naturalness remains a challenging problem. This paper proposes an image stitching method which preserves the flatness of planes in the scene for a natural look. Our method formulates the alignment of images as the camera parameters and the normal vectors of planes. Given a set of feature point matches, a process of grouping points into different layers and rejecting outliers is introduced. According to the epipolar constraint of corresponding points in two images, the focal length and the pose change of the camera are recovered simultaneously. Then, the normal vectors are estimated from the point pairs. To achieve good alignment and guide the warping of images, the model is combined with the mesh deformation as a global similarity constraint. In addition, bundle adjustment is adopted to maintain the consistency for stitching multiple images. Experiment shows that the proposed approach outperforms some state-of-the-art warps on real-world scenes.

In this study, a data-augmentation method is proposed to narrow the significant difference between the distribution of training and test sets when small sample sizes are concerned. Two major obstacles exist in the process of defect detection on sanitary ceramics. The first results from the high cost of sample collection, namely, the difficulty in obtaining a large number of training images required by deep-learning algorithms, which limits the application of existing algorithms in sanitary-ceramic defect detection. Second, due to the limitation of production processes, the collected defect images are often marked, thereby resulting in great differences in distribution compared with the images of test sets, which further affects the performance of detect-detection algorithms. The lack of training data and the differences in distribution between training and test sets lead to the fact that existing deep learning-based algorithms cannot be used directly in the defect detection of sanitary ceramics. The method proposed in this study, which is based on a generative adversarial network and the Gaussian mixture model, can effectively increase the number of training samples and reduce distribution differences between training and test sets, and the features of the generated images can be controlled to a certain extent. By applying this method, the accuracy is improved from approximately 75% to nearly 90% in almost all experiments on different classification networks.

In this paper, we consider the identification problem for nonlinear state-space models with skewed measurement noises. The generalized hyperbolic skew Student’s t (GHSkewt) distribution is employed to describe the skewed noises and formulate the hierarchical model of the considered system. A unified framework for estimating unknown states and model parameters is presented based on expectation-maximization (EM) algorithm, in which the forward filtering backward simulation with rejection sampling (RS-FFBSi) is employed to efficiently estimate the smoothing densities of the hidden states, and optimization method is adopted to update model parameters. One numerical study and the electro-mechanical positioning system (EMPS) are employed to verify the effectiveness of the developed approach.

This article develops a robust global strategy for identifying the linear parameter varying (LPV) errors-in-variables (EIVs) systems subjected to randomly missing observations and outliers. The parameter interpolated LPV autoregressive exogenous model with an uncertain/noisy input is investigated and a nonlinear state-space model is considered for the input generation model (IGM). The parameters estimation of the LPV EIV systems with nonideal observations is realized using the expectation–maximization algorithm which is particular effective for the incomplete data issue. To ensure the robustness in the identification, the Student’s t-distribution which is characterized by its adjustable degree of freedom, is used to handle the measurement non-normality. Since the posterior distributions of the latent states in the IGM are also involved in the identification process and they are difficult to calculate directly, the particle filter is introduced to recursively approximate them instead. Finally, the verification examples are given to demonstrate the effectiveness of the developed strategy.

Because of various problems in the production of ceramics, such as material and transportation, there exists many defects in ceramics products. Traditional methods use pixel-level segmentation to locate defects which only achieve pixel-level accuracy. In order to realize higher-precision defect localization, we propose a novel subpixel interpolation method based on the blurred edge model. Based on the analysis to the gray histogram of the defect image, the threshold segmentation method is first used to roughly locate the defect. The formula for calculation of subpixel position using interpolation method is then derived, which is faster than the traditional fitting method. Finally, the 2-D subpixel edge localization problem is disassembled 1-D calculation problems in the X-direction and Y-direction according to the direction of the defect edge gradient. The subpixel location calculation method is further given. Experiments on different defects show that the proposed method can precisely segment the defects.

The visual inspection algorithms are the core of automatic optical inspection system on surface mounting machines. This article is concerned with the development of precise positioning algorithms for circular mark points and transistor (TR) components on surface mounting devices. To handle nonuniform illumination or occlusion of other components, a polar coordinate transform and smoothness selection based circular mark point location method is proposed. The TR components are fundamental chips in electronic products and have various package types. The illumination changes, background disturbance, and the diversity of package types have imposed great challenges on the development of the uniform algorithm for detection and location of TR components. To deal with these issues, the 1-D integral image based TR component detection and location algorithm is proposed and the coordinates and orientation of the component are calculated simultaneously. The efficiency of the proposed methods is tested on real images and compared with classical Hough transform method, commercial algorithms on SMT482 device, and two methods of Halcon software.

A robust multiple model strategy for linear parameter varying (LPV) systems identification relied on Student's t-distribution is studied in this article. Since most industrial processes operate over several working points to meet multiple production objectives, a single model is generally insufficient to describe the process behavior over the whole operating range. To cope with this issue, the multiple model strategy is employed and at the same time, the outliers and missing measurements problems are both considered in this article. The finite impulse response (FIR) model is chosen to describe the local behavior of the LPV system. The statistical scheme based on the t-distribution is utilized to model the system noise so as to deal with the outliers. The problem of parameters estimation with incomplete outputs is addressed by using the expectation-maximization (EM) algorithm. The feasibility and effectiveness of the developed approach are proved via a mechanical unit.

Computer image processing technology has been widely used in ceramic product defect detection. However, there are few studies on automatic defect detection methods for ceramic products with complex surface geometry, such as ceramic wash basins. The main difficulty is that this type of sanitary ceramic product has a complex surface, strong light reflection, and very little texture information. It is hard to find a perfect illumination to make the defects obvious. In view of the above problems, this paper proposes a defect detection method based on the neighborhood pixel gray threshold, and conducts defect detection tests on the imaging results of ceramic wash basins with cracks on the surface under general light conditions, and obtains good results.

In this paper, we propose a modified CenterNet to complete the defect detection of Sanitary Ceramics. Generally, visual quality inspection is rather important during the productive process of Sanitary Ceramics and it is nearly impossible to inspect the massive images by hand. Consequently, it is necessary to devise an accurate and real-time system to process the data. However, due to the varied shapes and backgrounds of ceramics, conventional computer vision methods are usually not robust to all those variables. Detectors based on Deep Learning start to be adopted in recent years, but most algorithms require some carefully devised anchor boxes and post-processing methods, which also bring more computational costs. Here we decide to take advantage of the anchor-free model, CenterNet. We change the main structure to fit our own data and introduce an extra branch with shallow layers to strengthen the feature representation. The results have shown the great power of this model. Without even any post-processing methods, our model achieves a result of 96.16 AP on the established dataset.

The robust multimodel solution for multiple-input single-output nonlinear time-delay systems identification with polluted outputs is derived in this article. First, all the local autoregressive exogenous models are preidentified at the working points; then, the global system model is built by interpolating the local models with a smoothing strategy. The outliers and input time-delays which often increase the nonideality of process data are both considered. To cope with the outliers, the Laplace distribution is reutilized to describe the output measurement process and the negative impact resulted from each outlier imposed on parameter estimation can be suppressed through automatically assigned small weight. The parameter estimation procedure is realized with the expectation-maximization algorithm and the joint posterior probability of all input delays is also maximized to calculate the unknown input time-delays. With the verifications on a numerical example and the continuous fermentation process, the validity of the proposed approach is proved.

This article deals with the template matching problem, and a weighted smallest deformation similarity measure, which is robust to occlusions, background outliers, and complex deformations. The appearance-based nearest neighbor (NN) matching of points is constructed and the smallest location distance between each point in the template and its matching points is employed to penalize the deformation explicitly. Then, the weights are added to points in the template relied on their likelihood of belonging to the background through NN matching with the points around the target window. Experiments show that the proposed method improves the state-of-the-art performance on real-world scenario benchmarks and can be applied in rough positioning of surface mount technology components.

This paper investigates the robust identification of nonlinear systems in state-space setting with output measurements contaminated with outliers and part of output measurements missing at random. The problems of outliers and missing observations are often encountered in practical industrial processes and are taken into consideration comprehensively in this work. The robust Student's t-based observation model is built to model the output measurements with stochastic outliers, and the impact of outliers imposed on nonlinear system identification can be suppressed through inference of the degree of freedom in Student's t-distribution. The proposed robust nonlinear system identification method with an incomplete dataset is derived with the expectation-maximization (EM) algorithm, and the particle filter is employed to numerically approximate the Q-function of the EM algorithm. A numerical example and a chemical process are utilized to demonstrate the superiority of the proposed strategy.

This paper is concerned with the nonrigid point set registration problem and a probability-based registration algorithm with local connectivity preservation is proposed. A unified formulation for point set registration problem is introduced and the derived energy function is composed of three parts, distance measurement item, transformation constraint item, and correspondence constraint item. In order to preserve the local structure of point set, the definitions of -connected neighbors and connectivity matrix are given and the local connectivity constraint is constructed as a weighted least square error item. The point set registration problem is formulated in the expectation-maximization algorithm scheme and the optimal spatial transformation and correspondence matrix are estimated simultaneously. The effectiveness of the proposed method is verified by applying the method to synthetic point sets and real scenarios of hand shapes and surface-mount technology components.

Component pick-and-place technology has been widely used to improve production efficiency and reduce common defects. The vision-driven measurement system of a component pick-and-place machine requires an appropriate positioning algorithm with low computational complexity, high accuracy, and high generalizability. To satisfy these attributes is rather challenging. This paper focuses on the online component positioning problem based on corner points. Thus, we propose a robust, accurate, and efficient universal algorithm that incorporates preprocessing, coarse positioning, and fine positioning stages. Two types of model key points are introduced for interpreting the model component. To enhance positioning accuracy and robustness against illumination changes, the Harris corners and subpixel corner points are extracted from the images of real components. In the coarse positioning step, distance and shape feature matching methods are introduced to, respectively, compute the coarse and correct correspondences between type I model key points and Harris corner points. After the corresponding point pairs have been obtained, the coarse and fine positioning problems are formulated as least squares error problems. The effectiveness of the proposed method was verified by applying the method in several real component positioning experiments.

The robust parameter and output estimation for linear parameter varying (LPV) time-delay system with output data contaminated with outliers and subjected to randomly missing measurements are considered in this paper. The outliers, missing data, and the time-delay are widely existed in practical industry and have imposed extra difficulties on complex process modeling. The robust probability model to describe the LPV time-delay system is constructed with the student's t -distribution and the estimation problems are formulated in the framework of generalized expectation-maximization algorithm. The time-delay and parameter varying process properties, the outliers, and randomly missing measurements are taken into consideration comprehensively in the derivations of proposed algorithm and the unknown model parameters, scale parameter, degree of freedom parameter, the time-delay, and the noise-free output data are estimated simultaneously. The numerical example and a practical chemical process are used to present the efficacy of proposed algorithm.

This paper deals with the identification problem of a linear parameter varying (LPV) model with a noisy (uncertain) scheduling variable. An autoregressive exogenous (ARX) model is used to describe the system behavior and a nonlinear state-space model is used to model dynamics of the noisy scheduling variable. The uncertain scheduling variable is estimated by using a particle smoother, which is based on sequential Monte-Carlo (SMC) method, and all the unknown parameters in the LPV ARX model are identified under the framework of an expectation-maximization (EM) algorithm based on the intput-output data of the system. A numerical example and a mechanical unit are used to validate the proposed approach.

In this paper, a robust level set method is proposed for image segmentation. Traditional level set methods are sensitive to noise in images which greatly limits its application in real project. To overcome this shortcoming, the fractional order regularization and Markov random fields term are incorporated into the traditional level methods in this paper. The fractional order regularization can reveal more details of the image and the Markov random field (MRF) term takes the hole image into account. In additional to these two terms, a region term and a penalty term are added into the energy function. The comparison of the proposed method with the classical level set method is made and the results show that the proposed method is robust to noise in images in image segmentation application.

This study is concerned with the surface mount component positioning problem and an improved chamfer matching method based on iterative position and rotation angle estimation approach is proposed. Instead of performing matching of the image with different pre‐specified angle templates pixel by pixel, as does in traditional chamfer matching methods, the gradient information of the distance transform image is incorporated into the chamfer matching method and reduced computational cost of the method is achieved. The iterative formulas to update the translation and rotation angle are derived by reference to the characteristic of rigid body motion. The criterions of search region establishment, seed point selection and terminal conditions are given. The effectiveness of the proposed method is verified by applying the method to component positioning with actual captured images.

This paper addresses the problems of robust global identification and fast-rate output estimation for linear parameter varying (LPV) dual-rate systems with output measurements subjected to random time-delays and outliers in statistical framework. In practical industry, the process data are often dual-rate sampled, and the output data are usually contaminated with outliers and may be subjected to uncertain time-delays due to lab analysis, long-distance or network transmission, etc. The LPV dual-rate model is given and the robust global identification and output estimation problems are formulated in statistical scheme with the Laplace distribution. The robust identification algorithm to estimate all the unknown parameters and output data are derived in the generalized expectation-maximization algorithm framework and the random time-delays and outliers in output data are handled adaptively in identification process. The proposed algorithm is presented and verified through numerical simulation and a practical chemical process.