We introduce a smart glasses application to have museum video tour for one person who enjoys art. This application was developed in the 2020 smart glasses internship at Ricoh Company, Ltd.

View full abstract

We report on our participation to the Internship Program Of Fraunhofer FIT. It took place from February to March 2020. In the Internship, we discussed and studied the Retro-Fit which is an additional introduction of new technology to the existing factory equipment. We used the UbiFactory as equipment. As new technologies, we challenged to recognize the remaining quantity of materials in the warehouse by image processing and to visualize the equipment information by AR headset.

View full abstract

In this paper, the purpose is to share what I have learned and realizations through my internship and to deepen the reader's understanding of internships.

View full abstract

Spectral measurements are used in a material analysis and scene recognition . The conventional spectrometer has a trade off between the resolution and the measurement range. In this study, we propose a super resolution method with sensor tilting and spectrally varying blur kernel estimation from sparse atomic emission lines. We experimentally validated that a spectrometer with the tilted sensor could achieve about 20 times better resolution than the raw measurement

View full abstract

In removing unnecessary reflection components in the hologram mirror type imaging system, we proposed the removal of background components in color images using three-color HOE, and confirmed the effectiveness by simulation.

View full abstract

A part of the light transmitted through semitransparent liquids is absorbed by the liquids and its energy is attenuated. The change in light energy due to the absorption depends on the spectral absorption coefficients of liquids and the optical path lengths of light rays, and is described by the Lambert-Beer's law. In general, spectral absorption coefficients of liquids are often unknown in real-world applications, because they depend not only on liquid media themselves but also on dissolved materials. Accordingly, in this paper, we propose a method for estimating the spectral absorption coefficients of a semitransparent liquid from the hyperspectral images of objects immersed in the liquid. Specifically, our proposed method estimates spectral absorption coefficients on the basis of Non-negative Matrix Factorization, since the observed spectral radiance values, spectral absorption coefficients, optical path lengths, and Fresnel terms are all non-negative.

View full abstract

In this study, we propose a method to measure the spectral reflectance of silk fabric from an image and reproduce the color of silk fabric its using a reflection model. The spectral reflectance is only the diffuse reflection component must be measured, however in the case of silk fabric, the specular reflection due to the three-dimensional woven structure is incident, and the measurement accuracy of the spectral reflectance is reduced. In order to improve estimation accuracy, we propose a method to estimate the spectral reflectance of silk fabrics based on image measurement. Here, the image is measured with a camera equipped with a macro lens on the surface. Then, we propose a method of estimating the spectral reflectance from the diffuse reflection component by removing the specular reflection component from the histogram analysis of the measured image. The result shows that the accuracy of the estimation of the spectral reflectance of silk fabrics was improved.

View full abstract

We show how to construct a multispectral infrared imager and show how is can be used to perform real-time quantitative imaging of gas leaks. In order to make this possible, it is necessary to radiometrically calibrate of multiple microbolometer cameras, and we show how to make this possible. Together, these hardware achievements make autonomous gas leak safety systems possible for the first time.

View full abstract

When the viewing video contains screen shake information, countermeasures against VIMS are one of the important issues. So far, we have studied a screen shake extraction method using a motion histogram for screen shake. In this method, it is important to deal with false positives when we capture the change of histogram in the time direction. Therefore, in this paper, we consider the motion detection performance in the motion histogram generation method.

View full abstract

The plane estimation method used to detect VKOP feature points has multiple parameters, and it is necessary to fine-tune the parameters in order to obtain appropriate results. We propose a method for specifying appropriate parameters that does not require this fine adjustment.

View full abstract

Plane detection is a core task in 3D point clouds processing with numerous applications such as registration, object recognition, intelligent robotics, and SLAM. However, RANSAC and other methods based on random sampling are negatively affected by the inliers ratio of the point cloud, and the sliding voxel method depends on the voxel size to ensure its precision. Therefore, in this paper, we propose a multi-scale sliding voxel plane detector that can detect and segment the most reliable planes at each scale efficiently.

View full abstract