We propose a label enhanced and patch based deep learning phase retrieval approach which can achieve fast and accurate phase retrieval using only several fringe patterns as training dataset. To the best of our knowledge, it is the first time that the advantages of the label enhancement and patch strategy for deep learning based phase retrieval are demonstrated in fringe projection. In the proposed method, the enhanced labeled data in training dataset is designed to learn the mapping between the input fringe pattern and the output enhanced fringe part of the deep neural network (DNN). Moreover, the training data is cropped into small overlapped patches to expand the training samples for the DNN. The performance of the proposed approach is verified by experimental projection fringe patterns with applications in dynamic fringe projection 3D measurement.

The non-contact three-dimensional measurement and reconstruction techniques have played a significant role in the packaging and transportation of precious cultural relics. This paper develops a structured light based three-dimensional measurement system, with a low-cost for cultural relics packaging. The structured light based system performs rapid measurements and generates 3D point cloud data, which is then denoised, registered and merged to achieve accurate 3D reconstruction for cultural relics. The multi-frequency heterodyne method and the method in this paper are compared. It is shown that the relative accuracy of the proposed low-cost system can reach a level of 1/1000. The high efficiency of the system is demonstrated through experimental results.

This paper proposes a deep convolutional neural network (CNN) -based technique for the detection of micro defects on metal screw surfaces. The defects we consider include surface damage, surface dirt, and stripped screws. Images of metal screws with different types of defects are collected using industrial cameras, which are then employed to train the designed deep CNN. To enable efficient detection, we first locate screw surfaces in the pictures captured by the cameras, so that the images of screw surfaces can be extracted, which are then input to the CNN-based defect detector. Experiment results show that the proposed technique can achieve a detection accuracy of 98%; the average detection time per picture is 1.2 s. Comparisons with traditional machine vision techniques, e.g., template matching-based techniques, demonstrate the superiority of the proposed deep CNN-based one.

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 to registration errors, measurement errors, deformations, etc., a fast and accurate method is important in machine vision measurement. This paper builds a convenient and economic multi-view stereo (MVS) measurement system based on a linear stage and a rotary stage to reconstruct the measured object surface completely and accurately. In the proposed technique, the linear stage is used to generate the trigger signal and synchronize the laser sensor scanning; the rotary stage is used to rotate the object and obtain multi-view point cloud data, and then the multi-view point cloud data are registered and integrated into a 3D model. The measurement results show a measurement accuracy of 0.075 mm for a 360° reconstruction in 34 s, and some evaluation experiments were carried out to demonstrate the validity and practicability of the proposed technique.

Recent approaches based on convolutional neural networks significantly improve the performance of structured light image depth estimation in fringe projection and speckle projection 3D measurement. However, it remains challenging to simultaneously preserve the global structure and local details of objects for the structured light images in complex scenes. In this paper, we design a parallel CNN-transformer network (PCTNet), which consists of a CNN branch, a transformer branch, a bidirectional feature fusion module (BFFM), and a cross-feature multi-scale fusion module (CFMS). The BFFM and CFMS modules are proposed to fuse local and global features of the double branches in order to achieve better depth estimation. Comprehensive experiments are conducted to evaluate our model on four structured light datasets, i.e. our established simulated fringe and speckle structured light datasets, and public real fringe and speckle structured light datasets. Experiments demonstrate that the proposed PCTNet is an effective architecture, achieving state-of-the-art performance in both qualitative and quantitative evaluation.