Session Index

In this study, we propose a deep learning-based three-dimensional (3D) surface imaging method for specular and diffuse reflection composite surfaces. The proposed system combines artificial intelligence for 3D optical measurement integrated with digital holography and fringe projection technologies. 3D fusion by 3D segmentation and alignment for composite surfaces are driven by artificial intelligence processes. The experimental results show that the proposed method can achieve competitive performance on 3D surface imaging using the deep learning-based integrated coherent system.

  Preview abstract

We proposed a lensless encryption face recognition system based on deep learning. An amplitude-modulated mask is applied as an element for optical encryption. Then, a ResNet-50 architecture was trained to learn a well-structured representation space for the visual recognition task. This concept was validated by 10-class dataset with 90% sensitivity and 98% specificity, respectively. Such unconventional optic offers a variety of different possible setups.

  Preview abstract

In this study, we propose a deep learning-based three-dimensional (3D) surface imaging method for specular and diffuse reflection composite surfaces. The proposed system combines artificial intelligence for 3D optical measurement integrated with digital holography and fringe projection technologies. 3D fusion by 3D segmentation and alignment for composite surfaces are driven by artificial intelligence processes. The experimental results show that the proposed method can achieve competitive performance on 3D surface imaging using the deep learning-based integrated coherent system.

  Preview abstract

In this paper, we present a theoretical model for color breaking in a holographic-optical-element-based augmented reality (AR) display. The discussions about the design parameters of holographic optical elements (HOEs) for AR glasses, focusing on color uniformity as a function of the waveguide thickness, are conducted. This model can be used to evaluate the color breaking of a displayed white image from the user’s point of view. The simulation results show that color breaking occurs due to the limitations of pupil size and image shifting caused by the extended eye-box. Moreover, the thickness of the waveguide also causes uneven color distribution.

  Preview abstract

This research introduces a light-derive phase modulator based on dye-doped liquid-crystal (DDLC) films with a single-side rubbed cell. By using different polarized beams to stimulate azo dye, the alignment of the liquid-crystal demonstrates a change. The characteristic is harnessed for recording and erasing information. This optical element can be likened to a reusable computer-generated hologram (CGH).

  Preview abstract

The purpose of this study is to explore the integration of 3D printing of photo-polymerization with the polymer-dispersed liquid crystal (PDLC) technique to create 4D printing optical devices. The printed 3D elements, including Fresnel lenses and gratings, are combined with polymer-dispersed liquid crystals, which enable the printed 3D elements to possess the ability of varying light intensity and therefore exhibit the characteristics of 4D printing. This study demonstrates the possibility of simplifying the fabrication processes of
switchable optical devices, improving limitations of 3D printing, reducing production cost, enhancing manufacturing efficiency, and offering the potential of a new configuration of 4D printing.

  Preview abstract

This research is devoted to developing a portable and user-friendly 3D coordinates point cloud acquisition system that is cost-effective and resistant to external light interference. Unlike conventional techniques, this system is integrated with a single inertial measurement unit (IMU) and a planar optical displacement sensor for the 3D scanning. By measuring variations in tilt angles and displacement distances of the object, we can derive a rotation matrix that enables the conversion of a 2D displacement vector into a representation in 3D space. Finally, through the integration of minute coordinate variations for each dataset, the point cloud acquisition could be completed.

  Preview abstract

This study presents a method to improve the exposure uniformity of a hologram printer by using localized adapt modulation. By measuring the actual illumination of the projected light pattern, an adapt mask function, which can compensate the nonuniformity, is produced and incorporated into patterns to be displayed on the DMD. The results indicate that the uniformity of the projected light patterns is significantly improved.

  Preview abstract

This paper combines Otsu's binarization and accumulated intensity values to detect the installation arrangement of aluminum plates. The installation arrangement was the one of the 2023 intelligent robot competition topic presented by CHENBRO CO. and SYNTEC TECHNOLOGY CO (real industry requirement). Otsu's binarization can reduce the brightness influence in the environment. The accumulated intensity values are designed for the installation arrangement. In the experiments re-build by our laboratory (avoid the problem of product confidentiality), compared to YOLOv8 with the 80 images and running time 3 hours, the proposed algorithm effectively uses one image and running time 5 minutes.

  Preview abstract

This research focuses on spectral analysis and deep learning of air pollution images for PM2.5 and water quality images for BOD (Biochemical Oxygen Demand). A total of 150 air pollution images and 90 water pollution images were used for this study. These images were divided into training and testing datasets, and then fed into two neural network models for training. Finally, the accuracy of the model predictions was evaluated.

  Preview abstract

In this paper, a method for reconstructing the surface color of an object was proposed. It is based on the R, G, and B projected technology, and color theory regulated by the CIE color system, so it has the merits of simple optical structure and convenient operation, and quick analysis. In the experiment, a coin was used to verify the feasibility of the method, and the surface color has a good effect of restoring.

  Preview abstract

A 2D fringe pattern is designed and employed to perform 3D profile measurements by means of coaxial fringe projections and image acquisitions in short depth of field. To enhance the reliability and systematic accuracy, a signal processing algorithm is proposed as well. Accuracy of the retrieved 3D profile can be achieved in the order of sub-millimeters.

  Preview abstract