We propose a label enhanced and patch based deep learning phase retrieval approach which can achieve fast accurate using only several fringe patterns as training dataset. To the best of our knowledge, it is first time that advantages enhancement strategy for are demonstrated in projection. In proposed method, labeled data dataset designed to learn mapping between input pattern output part neural network (DNN). Moreover, cropped into small overlapped patches expand samples DNN. The...

Metal component surfaces are random textured and non-smooth. There many stains on the surface of metal that similar to gray scale scratches. The scratches have non-uniform distribution, various shapes, low contrast in their background, posing challenges accurate scratch detection. This paper presents a method for detecting weak based deep convolutional neural networks (DCNNs). First, DCNN is trained using labeled images. Then, some faults detected by DCNN, most can be removed through...

This paper proposes a deep convolutional neural network (CNN) -based technique for the detection of micro defects on metal screw surfaces. The we consider include surface damage, dirt, and stripped screws. Images screws with different types are collected using industrial cameras, which then employed to train designed CNN. To enable efficient detection, first locate surfaces in pictures captured by so that images can be extracted, input CNN-based defect detector. Experiment results show...

In optical metrology, state of the art algorithms for background and noise removal fringe patterns are based on space-frequency analysis. this Letter, an approach variational image decomposition is proposed to remove from a pattern simultaneously. method, directly decomposed into three components: first one containing background, second fringes, third noise, which described in different function spaces solved by minimization functional. A simple technical process involved algorithm improves...

The non-contact three-dimensional measurement and reconstruction techniques have played a significant role in the packaging transportation of precious cultural relics. This paper develops structured light based system, with low-cost for relics packaging. system performs rapid measurements generates 3D point cloud data, which is then denoised, registered merged to achieve accurate multi-frequency heterodyne method this are compared. It shown that relative accuracy proposed can reach level...

A new solution to the high-quality 3D reverse modeling problem of complex surfaces for fine workpieces is presented using a laser line-scanning sensor. Due registration errors, measurement deformations, etc., fast and accurate method important in machine vision measurement. This paper builds convenient economic multi-view stereo (MVS) system based on linear stage rotary reconstruct measured object surface completely accurately. In proposed technique, used generate trigger signal synchronize...

The weak symmetry relationship between the relative error and solution norm holds in our developed nonnegative least-squares truncated singular value decomposition method. By using this to specify optimal regularization parameters, we applied proposed algorithm recover particle size distribution from dynamic light scattering (DLS) data. Simulated results experimental validity demonstrate that method, which compliments CONTIN algorithm, might serve as a powerful simple approach inverse problem DLS.

The three-dimensional reconstruction of bubble flow field is great significance to study the motion gas-liquid two phase flow. Combined with abundant information bubbles in light images, a fast and accurate 3D method for dilute bubbly based on deep learning algorithm DIF-LeNet (Double Information Fusion LeNet5 net) proposed this work. calibration segmentation image data processing detailed. realizes fusion different dimension regression prediction. By DIF-LeNet, depth could be predicted by...

An early warning method of compressor valve fault based on multi-parameter signals (vibration, pressure, and temperature) is presented in this work. Due to the complexity working condition, run data are problems like noise feature aliasing, which makes it difficult extract useful features find out running law from original signals. In work, an improved deep learning network Multi-Level Fusion long short-term memory Component Evaluating Empirical Mode Decomposition Fuzzy C-Means (CEEMD-FCM...

In particle size measurement with dynamic light scattering (DLS), it is difficult to get an accurate recovery of a bimodal distribution (PSD) peak position ratio less than ~2:1, especially when large particles (>350nm) are present. This due the inherent noise in autocorrelation function (ACF) data and scarce utilization PSD information during inversion process. this paper, ACF investigated. It was found that initial decay section contains more information, for PSD. Based on this,...

In this paper, we are concerned with denoising in experimentally obtained electronic speckle pattern interferometry (ESPI) fringe patterns poor quality. We extend the application of two existing oriented partial differential equation (PDE) filters, including second-order single PDE filter and double filter, to ESPI very quality, compare them other efficient filtering methods, adaptive weighted improved nonlinear complex diffusion PDE, windowed Fourier transform method. All five filters have...

Dynamic light scattering (DLS) is one of the most popular techniques for particle size distribution (PSD) determination particles under Brownian motion. However, accurate retrieval PSD from bimodal with a peak position ratio less than 2:1 still an open problem in single-angle DLS measurement. In this letter, we proposed new weighted constrained regularization (CR) method to improve accuracy inversion data. The CR significantly improves various particles, and, first time, reasonably resolves...

Abstract Simulated and experimental data were used to investigate the particle size distribution (PSD) recovery from a flowing aerosol via dynamic light scattering (DLS). It is found that, for unimodal PSD, increasing flow velocity leads an increase in error of peak position, decrease height broadening PSD. This exacerbated larger particles. For bimodal separation two peaks decreases results single-peak PSD as increases. The effect on measurement similar adding signal noise, which increases...

A full-view three-dimensional (3-D) measurement method for complex surfaces is proposed, where 3-D data standard balls with different angles are used to calibrate the rotation axes of a turntable and obtain transformation matrices adjacent views. It can achieve accurate registration views obtains in conjunction principal point calibration cameras modified triple-frequency six-step phase-shifting phase demodulation methods. Experiments show that developed system based on proposed automatic...