Printed circuit board (PCB) inspection is an essential part of PCB production process. Traditional bare defect detection methods have their own defects. However, the method based on automatic optic a feasible and effective method, it having more application in industry. Based idea reference comparison this study aims at studying classification First all, extracting areas using morphology studied; meanwhile, data set containing 1818 images with 6 different detailed area image parts are...

To coupe with the difficulties in process of inspection and classification defects Printed Circuit Board (PCB), other researchers have proposed many methods. However, few them published their dataset before, which hindered introduction comparison new In this paper, we a synthesized PCB containing 1386 images 6 kinds for use detection, registration tasks. Besides, reference based method to inspect trained an end-to-end convolutional neural network classify defects. Unlike conventional...

To cope with the difficulties in inspection and classification of defects printed circuit board (PCB), many methods have been proposed previous work. However, few them publish their datasets before, which hinders introduction comparison new methods. In this study, HRIPCB, a synthesised PCB dataset that contains 1386 images 6 kinds is for use detection, registration tasks. Besides, reference-based method adopted to inspect an end-to-end convolutional neural network trained classify defects,...

Abstract The high-resolution (HR) spatio-temporal flow field plays a decisive role in describing the details of field. In acquisition HR field, traditional direct numerical simulation (DNS) and other methods face seriously high computational burden. To address this deficiency, we propose novel multi-scale temporal path UNet (MST-UNet) model to reconstruct spatial fields from low-resolution (LR) data. Different previous super-resolution (SR) model, which only takes advantage LR data at...

Since they can help people detect the early signs of diseases, accurate diagnostic techniques based on biomarkers are crucial in biomedical research. This article proposes a novel bivariate time-varying coefficients logistic regression model for addressing combined longitudinal biomarkers. Using B-splines method to estimate proposed model, we effectively combine multiple and improve accuracy. We show that is theoretically consistent. And it exhibits superior performance compared existing...

The physics-informed neural network (PINN) is effective in solving the partial differential equation (PDE) by capturing physics constraints as a part of training loss function through automatic differentiation (AD). This study proposes hybrid finite difference with PINN (HFD-PINN) to fully use domain knowledge. main idea finite-difference method (FDM) locally instead AD framework PINN. We at complex boundaries and FDM other domains. To avoid background mesh, we propose HFD-PINN-sdf, which...

Simultaneous localization and mapping (SLAM) is one of the most essential technologies for mobile robots. Although great progress has been made in field SLAM recent years, there are a number challenges dynamic environments high-level semantic scenes. In this paper, we propose novel multimodal system (MISD-SLAM), which removes objects reconstructs static background with information. MISD-SLAM builds three main processes: instance segmentation, pixels removal, 3D map construction. An...

Physics-Informed Neural Networks (PINNs) can be regarded as general-purpose PDE solvers, but it might slow to train PINNs on particular problems, and there is no theoretical guarantee of corresponding error bounds. In this manuscript, we propose a variant called Prior Dictionary based (PD-PINNs). Equipped with task-dependent dictionaries, PD-PINNs enjoy enhanced representation power the tasks, which helps capture features provided by dictionaries so that proposed neural networks achieve...

There have been several efforts to Physics-informed neural networks (PINNs) in the solution of incompressible Navier-Stokes fluid. The loss function PINNs is a weighted sum multiple terms, including mismatch observed velocity and pressure data, boundary initial constraints, as well residuals equations. In this paper, we observe that combination competitive functions plays significant role training effectively. We establish Gaussian probabilistic models define where noise collection describes...

Recently, unsupervised methods for monocular visual odometry (VO), with no need quantities of expensive labeled ground truth, have attracted much attention. However, these are inadequate long-term task, due to the inherent limitation only using data and inability handle error accumulation problem. By utilizing supplemental low-cost inertial measurements, exploiting multi-view geometric constraint sequential constraint, an visual-inertial framework (UnVIO) is proposed in this paper. Our...

Temperature monitoring during the life time of heat source components in engineering systems becomes essential to guarantee normal work and working these components. However, prior methods, which mainly use interpolate estimation reconstruct temperature field from limited points, require large amounts tensors for an accurate estimation. This may decrease availability reliability system sharply increase cost. To solve this problem, develops a novel physics-informed deep reversible regression...

AbstractPhysics-informed neural networks (PINNs) have been proposed to solve two main classes of problems: data-driven solutions and discovery partial differential equations. This task becomes prohibitive when such data is highly corrupted due the possible sensor mechanism failing. We propose Least Absolute Deviation based PINN (LAD-PINN) reconstruct solution recover unknown parameters in PDEs – even if spurious or outliers corrupt a large percentage observations. To further improve accuracy...

Physics Informed Neural Network (PINN) is a scientific computing framework used to solve both forward and inverse problems modeled by Partial Differential Equations (PDEs). This paper introduces IDRLnet, Python toolbox for modeling solving through PINN systematically. IDRLnet constructs the wide range of algorithms applications. It provides structured way incorporate geometric objects, data sources, artificial neural networks, loss metrics, optimizers within Python. Furthermore, it...

Intelligent forecasting of economic indexes has been an important demand for sustainable management smart cities. Existing methods this purpose were mostly established upon the basis mechanism. Econometric models are most general technical means in area. However, era digital economy, increasing amount big data brought great change to traditional production. It is becoming more difficult conventional technological deal with multi-dimensional indexes. To such challenge, paper introduces...

The rapid analysis of thermal stress and deformation plays a pivotal role in the control measures optimization structural design satellites. For achieving real-time satellite motherboards, this paper proposes novel Multi-Task Attention UNet (MTA-UNet) neural network which combines advantages both Learning (MTL) U-Net with an attention mechanism. Furthermore, physics-informed strategy is used training process, where partial differential equations (PDEs) are integrated into loss functions as...