Non-contact displacement measurement is necessary for evaluating the health of infrastructure structures. One of the non-contact displacement measurement technologies is the sampling moiré method. This method is capable of measuring displacements greater than the grating pitch by using time-series displacement information. However, if the displacement between one frame exceeds half the grating pitch or if there is an obstacle such as a person between the camera and the grating, the displacement cannot be measured correctly. In this paper, we propose a concentric moiré method in which the grating is replaced by concentric circles. This method makes it possible to measure displacements greater than the grating pitch using only spatial information rather than time series. We confirm that the proposed method makes it possible to measure displacements greater than half the grating pitch between one frame.

This study proposes a method for change detection by extracting small image pairs from recorded video pairs in patrol inspections of oil plants using a mobile robot equipped with a camera. While change detection methods often perform well in controlled laboratory environments, their performance tends to degrade in plant environments. This is because, in plant environments, the regions of change caused by anomalies within the images are often small, and the environments are structurally complex. In this study, to address this issue, we adopt an approach that extracts small image pairs from the target and reference images for change detection. Additionally, we develop a method for extracting small image pairs that considers the consistency of three-dimensional coordinates based on multi-view stereo. As a result, the proposed method achieved high-precision change detection, demonstrating its potential for application in patrol inspections of oil plants using a mobile robot.

In this study, an ortho image was generated from the 3D point cloud data of cherry tomatoes, and fruit regions were extracted using SegNet, a deep learning based semantic segmentation model. From the extracted fruit regions, a reconstructed 3D point cloud of each fruit was generated, and the occluded regions were complemented based on the fruit's geometric symmetry. The fruit volume was then estimated using the least-squares ellipsoidal approximation.

When the fruit was not occluded by leaves, the estimation error was within ±1 ml, demonstrating the effectiveness of the proposed method. In contrast, when the fruit was partially occluded by leaves, the error increased significantly to 18.66±14.33 ml, due to incomplete reconstruction and poor shape fitting with an ellipsoid. To address this issue, a novel method was introduced to estimate the fruit's center and symmetrically complement occluded regions. As a result, the accuracy of the volume estimation improved substantially, reducing the error to 1.73±3.91 ml.

Transfer learning is the one of methods to improve the efficiency of training for reinforcement learning (RL) agents. That is the method of transferring knowledge which is acquired in the past tasks or given by experts to similar tasks. However, if the RL agents don't retain knowledge that is effective for improving learning speed or quality, there is a problem in that it causes negative transfer, which reduces learning efficiency. This study proposes a method for deleting knowledge which are less the number of state observation through trial and error by RL agents. Our proposed method was applied to path planning for a two-wheeled mobile robot to verify its effectiveness, and it was confirmed that the proposed method reduced the amount of knowledge by 98% and the learning time by 32%.

Anatomical differences in the cortex between people with Major Depressive Disorder (MDD) and healthy subjects have been previously reported, and there are also functional differences of brain gamma-band activity between them. However, the effects of the anatomical differences on brain activity are still unknown. Gamma oscillation is considered as a region specific activity. As such, we constructed a personalized meso-scale brain simulation with the neural mass model that has biologically plausible parameters. Then we analyzed the differences of the synchronization of gamma-band neural activity in the cortex caused by the differences in the gray matter structure. We also generated the EEG signal from simulated neural activity and examined its appropriateness for simulating EEG values by comparing the generated signal with the simulation from the previous multi-compartment model. Finally, the simulated signals were compared between subjects with MDD and those of a healthy participant.

When humans experience stress, they exhibit a physiological coping responses, which is typically accompanied by an increase in blood pressure. This responses can be classified into three categories based on hemodynamic indices (mean blood pressure, cardiac output, and total peripheral vascular resistance) as either active coping responses, passive coping responses, or no stress coping. Conventional methods for labeling facial thermal images according to changes in hemodynamic indices often do not account for the delayed responses between skin temperature changes and hemodynamic activity. In this study, we employed t-SNE, a dimensionality reduction technique, to map high-dimensional facial thermal image data into a low-dimensional space. This approach enabled the simultaneous assessment of differences between facial thermal images and hemodynamic indices, classification of facial thermal images according to stress coping responses categories, and development of a stress coping responses discrimination model. The results indicate a temporal delay between facial thermal responses and hemodynamic indices, with the discriminant model achieving an accuracy of 81.3%, which surpasses the accuracy of conventional methods.

Interest in next-generation power grids, such as smart grids and distributed networks, has grown due to the goal of achieving carbon neutrality. Renewable energy sources, like solar power, are key to this transition, but their inherent fluctuation poses challenges. Our research focuses on a Virtual Grid (VG) system, an autonomous, distributed network that utilizes renewable energy. The VG system connects multiple VG Hubs to synthesize and distribute direct current power. Ensuring the effective control of each VG Hub amid network fluctuations is essential for the success of the system. Our research aims to develop an algorithm that determines power sources and routes for new loads within the VG Hub network. Previous methods to optimize the overall flow for each load addition have encountered issues with configuration costs and deviations from optimal solutions. Our challenge is to create a hybrid method that balances configuration minimization with flow optimization. The proposed algorithm reduces the need for new configurations with each new load and performs flow optimization only if the flow optimality of the solution exceeds a predefined threshold. Simulations were conducted to test the effectiveness of the algorithm in various network scenarios. The results demonstrated that our algorithm excels in networks with a high average degree and in large-scale networks with a low average degree. In addition, the algorithm performs better in scenarios with multiple supply sources or when additional power sources are introduced, compared to scenarios with a single supply source. By focusing on the correctness of our methods, the clarity of our objectives, the engagement of our solutions, and the effective delivery of our results, our approach presents a significant advancement in the control and optimization of Virtual Grid systems.

Building datasets is one of the most tedious tasks in machine learning. Image datasets consisting of unique images that are not included in large datasets require vast annotated images and bounding boxes that are created manually. We address this issue using a self-supervised learning (SSL) model for clustering before labeling the images. SSL has generally been used for pre-training. We propose the use of SSL to build image datasets by clustering using k-means. The dataset that we mainly used was CIFAR-10, and the SSL model that we used was Image Bidirectional Encoder Representations from Transformers (BERT) Pre-Training with Online Tokenizer (iBOT).

Computations in science and engineering have been becoming tough. Such computations are often outsourced to subcontractors so that users could concentrate on their main tasks (making models, discussion and so on). However, they must reveal important data to subcontractors. As a result, the data might be leaked. In order to prevent this security threat, many researchers have studied secure computation. It enables users to outsource without revealing important data. Fully homomorphic encryption (FHE) is a key technology to realize secure computation. This paper proposes a new extension of the IH method in FHE. The IH method is a method which enables users to encrypt data which is used for computing polynomial expressions by using a special matrix. While the usefulness of the IH method, its application is limited to polynomial expression: the case that the target computation is given as a polynomial expression. Now, this paper first proposes its extension to the case of rational expression. We find a newly revised matrix to be used in decryption even if the computation includes the operation of division and apply it to overcome the problem.

Fundamental properties of neuronal networks at the single-cell level can be understood by comprehensively capturing the activity of all neurons within the network. In this study, we constructed micro-networks on a high- density micro electrode array (HD-MEA) and performed electrical measurements of all neurons using a densely packed electrode configuration. We mapped the activity of individual neurons with cell positions on stained images.

In software development, debugging requires a lot of human resources. A breakthrough in the research of automatic bug fixing was made by GenProg, which uses genetic programming for fixing bugs. This study proposes automatic bug fixing by optimization methods that are computationally less expensive than genetic programming. As the evaluation experiments, buggy programs to find leap years was fixed with PSO and random. As a result, it was confirmed that PSO can modify the program faster than random.