Search

In general, manual methods are often used to inspect defects in the production of metal screws. As deep learning shines in the field of visual detection, this study employs the You Only Look At CoefficienTs (YOLACT) algorithm to detect the surface defects of the metal screw heads. The raw images with different defects are collected by an automated microscopic camera scanning system to build the training and validation datasets. The experimental results demonstrate that the trained YOLACT is sufficient to achieve a mean average accuracy of 92.8 % with low missing and false rates. The processing speed of the trained YOLACT reaches 30 frames per second. Compared with other segmentation methods, the proposed model provides excellent performance in both segmentation and detection accuracy. Our efficient deep learning-based system may support the advancement of non-contact defect assessment methods for quality control of the screw manufacture.

The three-dimensional (3-D) micro-scale surface imaging system based on the digital fringe projection technique for the assessments of microfiber and metric screw is presented in this paper. The proposed system comprises a digital light processing (DLP) projector, a set of optical lenses, a microscope, and a charge coupled device (CCD). The digital seven-step fringe patterns from the DLP projector pass through a set of optical lenses before being focused on the target surface. A set of optical lenses is designed for adjustment and size coupling of fringe patterns. A high-resolution CCD camera is employed to picture these distorted fringe patterns. The wrapped phase map is calculated by seven-step phase-shifting calculation from these distorted fringe patterns. The unwrapping calculation with quality guided path is introduced to compute the absolute phase values. The dimensional calibration methods are used to acquire the transformation between real 3-D shape and the absolute phase value. The capability of complex surface measurement for our system is demonstrated by using ISO standard screw M1.6. The experimental results for microfiber with 3 μm diameter indicate that the spatial and vertical resolutions can reach about 3 μm in our system. The proposed system provides a fast digital imaging system to examine the surface features with high-resolution for automatic optical inspection industry.

The fringe projection profilometry with sinusoidal patterns based on phase-shifting algorithms is commonly distorted by the nonlinear intensity response of commercial projector. In order to solve this issue, sinusoidal width modulation is presented to generate binary sinusoidal patterns for defocusing the projection. However, the residual errors in the phase maps are usually notable for highly accurate three-dimensional shape measurements. In this paper, we propose the fringe patterns of the sinusoidal, square, and triangular periodic waveforms with seven-step phase-shifting algorithm to further improve the accuracy of three-dimensional profile reconstruction. The absolute phase values are calculated by using quality guided path unwrapping. We learn that by properly selecting fringe patterns according to the target shape, the undesired harmonics of the measured surface have negligible effect on the phase values. The experiments are presented to verify the imaging performances of three fringe patterns for different testing targets. The triangular fringe patterns are suitable for the shape measurements of complex targets with curved surfaces. The results provide a great possibility for high-accuracy shape measurement technique with wider measuring depth range.

The characterization of surface micro-roughness is investigated by using off-specular measurements of polarized optical scattering. In the measurement system, the detection angles of optical scattering are defined by the vertical and level scattering angles. The rotating mechanism of angles is controlled by stepper motors. Waveplate and polarizer are used to adjust light polarization and detection. We conduct the optical scattering measurements by using four standard metal sheets of surface roughness. The nominal values (Ra) of standard micro-roughness are 1.6 μm, 0.8 μm, 0.4 μm, and 0.1 μm, respectively. Samples with different surface roughness are evaluated with the utilization of laser sources at three incident wavelengths. These polarized images are analyzed using a computer program to obtain the distribution of light intensity. The results show great correlation between the metal surface roughness and polarization states. This measurement system can be used to quickly and accurately distinguish between different surfaces and properties.

In this work, we thoroughly investigate the shape, size, and location of the photonic nanojets (PNJs) generated from the illuminated dome lens. The silk fiber is directly extracted from the cellar spider and used to form the dome lens by its liquid-collecting ability. The solidified dielectric dome lenses with different dimensions are obtained by using ultraviolet curing. Numerical and experimental results show that the long PNJs are strongly modulated by the dimension of the dome lens. The optimal PNJ beam shaping is achieved by using a mesoscale dielectric dome lens. The PNJ with a long focal length and a narrow waist could be used to scan over a target for large-area imaging. The silk fiber with a dome lens is especially useful for bio-photonic applications by combining its biocompatibility and flexibility. [ABSTRACT FROM AUTHOR]