We apply an existing reliability estimation method to vision-based self-localization in a RoboCup environment. The existing method has been used for LiDAR-based self-localization to measure the reliability of the estimation. For the judgement, a convolutional neural network (CNN) rates the consistency between the measurements from a LiDAR and an occupancy grid map. We replace the LiDAR and the map with a monocular color camera and a line map on a soccer field respectively. Our implementation was evaluated on a simulated soccer field. In the evaluation, it output appropriate evaluations toward the errors of vision-based Monte Carlo localization.

We have proposed a method for path-tracking of autonomous mobile robots by end-to-end learning of vision and action. Our method aims to diversify the navigation methods for these robots. It involves collecting a dataset while the robot moves and using this for learning. However, the density of data collection has not been discussed, and the density is uniform. This may result in excessive data collection in certain route areas. Addressing this, we first check for any bias in the collected dataset. We then propose a method to adjust the data collection density by varying the robot’s translational speed based on the training data. The effectiveness of our proposed method is confirmed through experiments in a simulator.

In the field of construction industry, illuminance measurement is expected to be automated using mobile robots. However, the present illuminance measurement robot is not able to execute the autonomous driving to the end. This is mainly caused by the distortion which emerges on the environment map. In this paper, we propose two methods to prevent the distortion on the environment map. The first method is a way to check the environment map on real time by comparing with the architectural drawing as a scale and eliminate the distortion if detected. The second method is a way to improve the fitness function used in the evolutionary computation for localization. Our results show that our methods can prevent the distortion on the environment map and contribute to stabilize the autonomous driving of the illuminance measurement robot. We conclude that our method is a promising technique to accomplish the illuminance measurement autonomously.

To promote the utilization of service robots, it is essential to establish a robot-friendly environment. Fiducial markers serve as one of the tools to assist in localization. We have developed high-accuracy fiducial markers with over ten times the measurement accuracy of conventional 2D fiducial markers. In this paper, we describe the structure, measurement principles, and performance of this high-accuracy marker called LeagTag. Additionally, we present examples of its application in precise navigation tasks such as elevator door passage with an autonomous mobile robot, demonstrating its utility as a tool for creating robot-friendly environments.

The paper describes localization of autonomous mobile robot using partitioned 3D map. Partitioned 3D maps are obtained by partitioning the original whole 3D map. The experiments with a mobile robot are performed in indoor and outdoor environments. The experimental results of calculation cost for localization are shown with the whole 3D map and partitioned 3D maps.

Localization using radio waves is an essential technique for estimating communicating device locations. We are working on position estimation using the carrier phase as a fingerprint instead of radio wave strength. Carrier waves can be measured with higher accuracy than radio wave strength, so position estimation can be made behaviourally by using carrier waves. However, it has been difficult to measure the carrier phase. This paper proposes a system for constructing carrier phase maps using a communication device equipped with wireless bi-directional synchronization technology and a mobile robot. Using the proposed method, we have confirmed that it is possible to construct a carrier phase map for an indoor area of 2x3 m. The carrier phase map is expected to be applied to high-precision position estimation and action planning considering radio wave quality.

Smartphones are equipped with sensors such as GNSS (Global Navigation Satellite System) and IMU (Inertial Measurement Unit), and recently APIs have been released to access their raw sensor data. The performance and accuracy of using smartphones as sensors for robot navigation, especially for position estimation, has not been fully evaluated. Therefore, this paper compares GNSS and IMU observation data acquired by smartphones with commercially available GNSS receivers and IMUs. Comparative experiments in a static environment showed that the noise of the GNSS pseudorange observations was greater than that of commercial GNSS receivers, and the stability of the IMU accelerometer and angular rate sensor bias was comparable to that of commercial low-cost IMUs. These results provide guidelines for the use of smartphones as sensors for robot navigation.

This study aims to construct maps of ground and airborne risks for unmanned aircraft traffic management from aerial photo orthographic maps. Essentially, the ground risk is the possibility that an unmanned aircraft may harm people, and airborne risk is the possibility of collisions between aircraft. From this, ground risk was classified into levels based on the identification of objects on the ground, and airborne risk was expressed as the frequency of passing through a randomly planned route, resulting in the creation of ground and airborne risk maps. The Vision Transformer (ViT) was used to identify the ground risk map. The A* algorithm was also used for random route planning for aerial risk. As a result, the locations where traffic is likely to be particularly concentrated, and their ground conditions were visualized, and the aerial risk was identified when operating many unmanned aircraft with low ground risk.

This report presents a portable LiDAR SLAM and localization system that integrates an IMU. The system utilizes Generalized ICP for scan matching and employs a hierarchical geometric observer for state updates, including IMU velocity and biases. We evaluate the system using publicly available datasets recorded with a portable LiDAR. Results demonstrate that our system can accurately estimate sensor ego-motion and effectively map environments.

Controlling personal mobility (PM) in a narrow space is not easy. In particular, passing through a door is quite difficult because it requires simultaneous operation of door opening/closing and PM maneuvering. In this study, we propose a semiautomatic operation support system for passing doors by sharing roles, i.e., the door opening/closing by the operator and PM maneuvering by the system, to reduce the burden on the operator. The system enables PM to pass through doors regardless of the position of the door knob and the angle of PM toward the door while avoiding surrounding obstacle. The results of experiments indicates that the proposed semiautomatic system could pass through various doors more easily and comfortably than manual operations.

As the demand for personal mobility (PM) increases, studies on autonomous control systems enabling efficient and safe movement are advancing. In this study, we developed a semi-autonomous control system that allows for flexible switching between automatic and manual driving and enable seamless indoor/outdoor driving. The system achieves autonomous driving using global and local path planning, and autonomous navigation in indoor/outdoor environments through partitioning and floor movement detection. To enable an operator intervention, the system incorporated sub-systems for switching between automatic and manual drivings by button inputs, and augmenting the output of the automatic system through joystick operation (override). The results of the subject experiments confirmed that the proposed system was effective for comfortable mobility.

In this study, we address the development of a driver assistance system for electric kickboards aimed at preventing falls from steps, such as sidewalks, encountered when maneuvering to avoid pedestrians. Central to our investigation is the differentiation between instances where the driver is cognizant of the step and those where the step is unnoticed, positing that assistance is only requisite in the latter scenario. To facilitate this, we collect and analyze driving data capturing vehicle momentum under both conditions to compile a dataset. This dataset serves as the foundation for training an anomaly detection-based machine learning model, designed to estimate the driver's awareness of steps. Our approach offers a novel contribution to enhancing safety features in electric kickboard mobility, leveraging data-driven insights to anticipate and mitigate potential fall risks associated with urban navigation.

In recent years, the development of Personal Mobility Vehicles (PMV) has played a pivotal role in advancing people's right to mobility and introducing innovative modes of transportation. With the evolution of PMV, technologies such as follow-up driving and automatic driving have come to the forefront. This study explores the implementation of a centralized control system for PMV, aiming to significantly improve safety, comfort, and operational efficiency. Our research delves into the methodology for integrating a centralized control system, examining its potential impacts, challenges upon deployment, and its compatibility with user experiences. The effectiveness of the centralized control system will be evaluated through comprehensive driving test analyses.

Recently, Ultra-compact electric vehicles are now on sale that use electric motors as their driving source, which have less impact on the global environment. This vehicle has low rigidity on the outside. Therefore, road and wind noise generated while driving become interior noise. To solve this problem, we have proposed an active noise control system for ultra-compact electric vehicles using boundary vibration by giant magnetostrictive actuator. In the proposed system, the giant magnetostrictive actuator generates a high output using boundary-wall surface vibration. In this study, we developed a model of the giant magnetostrictive actuator and analyzed, by electromagnetic field analysis, output performance. From the result, in the development of giant magnetostrictive actuator for ultra-compact EV, we found that the selection of permanent magnets and giant magnetostrictive material is important to ensure stable output in the road noise frequency band.

In recent years, various types of personal mobility have been researched and developed. Among them, there is a seated-type personal mobility such as a wheelchair. The human body motion interface has developed for such type of the personal mobility. Although this interface uses body movements associated with voluntary movements, and the potential for intuitive operation has been confirmed, the lack of force feedback makes it difficult to correlate the body inclination with the wheelchair's speed. Therefore, this study proposes a belt type interface that adds force feedback to the human body motion interface of seated type personal mobility. This interface uses the tension of the seat belt to provide feedback to the user, thereby improving the feel of operation. In particular, this paper reports on an active seat belt system for this purpose.

In recent years, a “declining birthrate and aging population” is believed to cause a chronic shortage of labor in the fields of medical care and caregiving. Therefore, manual wheelchairs and electric wheelchairs are commonly used as means of mobility in medical or caregiving facilities and daily mobility assistance for pedestrians. However, the disadvantages of manual wheelchairs include the strain on the arms when moving, and the risk of falling and the difficulty of carrying an electric wheelchair. Therefore, to solve these problems, this research proposes incorporating a differential mechanism into an active castor that has an omni-directional movement function to improve the operating rate of the motor. Additionally, by mounting a new active castor equipped with a suspension function on a manual wheelchair in a detachable state. To verify that the manual wheelchair is electrically powered, we conducted verification of straight-line propulsion and on the spot turning drive methods.

Personal vehicles with high degrees of mobility are required to improve usability and safety. In this study, we propose a personal vehicle that can move not only longitudinally but also laterally by using a friction-drive active omni wheel (AOW). To solve the problems of control and slippage in the conventional friction-drive AOW, a reduced motor drive mechanism and a newly designed AOW are introduced, which contribute to reducing the size and weight of the drive mechanism of the personal vehicle. Experiments were conducted with prototypes of the proposed AOW and personal vehicle, and it was confirmed that they worked as expected.

Japan have one problem that the decline of the public transportaion by the aging society, so that the government of Japan have recommended the introduction of personal mobility vehicles (short for PMV) that one of way to solve. However, there are few PMVs considered the elderly people on the market, such as these vehicles have the problem to trip over the step. In this study, we developed the compact personal mobility vehicle equipped with the suppression unit and quantitatively evaluate its fundamental performance through verification experiments.

Japan have one problem that the environmental and energy problems, so that the government of Japan have recommended the introduction of electric vehicles (short for EV) that one of way to solve. However, the study for EV has focused on four-wheeled vehicles, so that the electric motorcycle has not studied much less in various research institutes. In this study, we developed the traction control system for two-wheel drive electric motorcycle and quantitatively evaluate the fundamental performance through verification experiments.

One of the social roles of aquariums is to create opportunities to think and learn about the ecology and living environment of species. We are developing a tactile augmented reality (AR) device that reproduces the sense of touch of aquatic organisms. In this paper, we have conducted basic experiments to build the system and to investigate the frequency characteristics of the object presenting the tactile sensation. As a basic experiment, we compared the vibration characteristics of a human finger and a measuring rod when touching an object.

We have developed a bellows-type vacuum suction pad with strain gauges attached to its sides. The strain gauges are placed and pasted in all four directions of the suction pad, so the pad operates as a six-axis load cell in its own right. The resistance value changes as the suction pad deforms, and this is read to detect compression, tension, and/or tilt of the suction pad. The strain gages were formed using a polyimide film as a substrate, on which a NiCr thin film was deposited by a sputtering method. The strain gauges were patterned using a shadow mask with gaps in the shape of the strain gauges during sputtering. This paper describes the principle of operation of the vacuum suction pad with strain gages and details of the fabrication method.

This paper proposes the Universal Dense Tactile Sheet. It is a distributed tactile sensor sheet which is wire-saving, multi-contact detectable, and dense enough to be applied to robot fingers universally. This sheet was achieved by densifying the conventional universal tactile sheet which is a flexible printed circuit able to be bent and cut according to each installation surface. The densification was accomplished by introducing the concepts of cutting between sensor cells and assembling PCB on the basis of each way of sheet installation. We confirmed the proposed sheet can be installed on two different real robot fingers without compromising its sensing functionality.

This study analyzes the recognition process of examining the texture of fine particles through tactile sensing using human fingers and aims at establishing the structure of texture recognition. Our previous study confirmed the occurrence of hysteresis in tactile sensation in the mid-teens. This study investigated the effect of the delay in sample presentation to the subject on the KANSEI evaluation of tactile sensation. The results showed that the KANSEI evaluation changed with the delay in the sample presentation.

Recently, fMRI is essential technology for cognitive neuroscience aiming to understand a relationship between brain activity and tactile sensation. However, most measuring instruments and experimental devices cannot be brought in the MRI room because high magnetic field is generated. In this study, to develop a tactile device that can be used in high magnetic field, an ultrasonic transducer using only materials with low magnetic permeability as a driving source of the tactile device was fabricated. Also, fMRI-compatibility tests were conducted closed to 3T MRI environment. The influences from the MRI environment to the transducer and those from the transducer to the MRI environment were investigated.

We propose tactile estimation system by pressure distribution sensors. When a fabric actuator is used as a tactile transmission device, the pressure distribution measured at the reader side should be able to be reproduced at the follower side ideally. However, this is difficult to achieve with the current control resolution of fabric actuators. we aim to solve this problem by creating a system to estimate tactile sensation and transmitting the tactile indirectly as the tactile label. The results showed that for three subjects, the estimation performance exceeded 80%, except one tactile. One tactile sense was not correctly estimated in one subject because of the variety of contact with the same label.

We have developed a tactile sensor module for sensing and controlling the degree of contact between a wearable device and a human. In our previous work, there were issues with the piping connection method to the tactile sensor module and the stretchable function of the clothing-type device that carries the sensor. In this paper, we describe a tactile sensor module and a clothing-type device that solve these issues.

In this study, we propose a skill transmission system using force-haptic projection mapping and presenting vibrotactile sensations. The system extends the conventional force visualization system by mapping haptic information as perceptual intensity. The system also provides feedback on the sensory experience during work using a sensory equivalent conversion method of high-frequency vibration. We believe that this system can be applied to the learning of skills that require delicate techniques by providing visual and haptic feedback on the physical sensations that are difficult to convey by words alone.

Stereohaptic vibration is an innovative vibrotactile technology that extends the conventional phantom sensation to the surrounding space, representing a virtual vibration source in the external environment. This vibrotactile technology is like a tactile version of auditory stereoacoustics, and since the interaction between stereoacoustics and vision has been reported, this technology may also interact with and reinforce each other with vision and hearing. In this report, we investigate the relationship between stereohaptic vibration and peripheral vision by presenting tactile information related to objects that appear in the peripheral vision and measuring the response speed to the objects.

We have proposed a bracelet-type tactile device that measures and presents vibrations transmitted to the wrist. In the previous paper, we proposed a method to compensate for individual differences in the mechanical properties of the fingers. In this paper, we first investigated the influence of force on the proposed method. We also plan to verify the influence of the individual difference correction when reproducing other people’s movements based on physical sensation.

Position-based visual servoing has the disadvantages of low initial calibration accuracy and sensitivity to noise such as environmental changes. Therefore, in order to enable highly accurate object tracking of position-based visual servoing considering disturbances, this study performs online calibration using the least-squares method during the operation of position-based visual servoing considering errors that occur during feature point extraction on images and calibration errors of the camera. The camera pose is calibrated using the least squares method during operation. This enables the position-based visual servoing to reach objects with high accuracy, considering the measurement errors.

Persistent homology, which captures topological features of point clouds and has been used successfully in the field of topology in mathematics. This approach can extract shape features robustly by low-dimensional (i.e., 3D in this paper context) points. We applied this technique to 3D point cloud processing in robot vision. Persistent homology captures holes in point cloud as pairs of birth time and death time of changing radius of point distance. By selecting these pairs by criteria based on the focusing hole parameter, corresponding hole can be detected. In this paper, we demonstrate that a method called inverse analysis of persistent homology can be used to robustly detect holes in industrial parts.

In this paper, we propose the ultra-thin optical camera system using metalens and an image sensor unit with suction for robotic assembly. In the system, the image is capable of two difference work distance of near and far by one image sensor. The unit can capture a target in the near view area while suctioning, and other targets in far view area while grasping the target. We prototyped the sensor unit with the system and the suction. In the experiment, the capability of imaging by metalens in the proposed configuration and the performances of suction function were confirmed. Moreover we demonstrated insertion of a board to board connector using a robot arm with our unit attached. The experimental results and the demonstration show the performance and the effectiveness of the system and the unit.

User calibration required to calculate user-specific parameters is time-consuming in eye-tracking. Therefore, implicit user calibration, which does not use explicit reference points, has been studied. However, typical calibration methods are limited to the display screen. In this paper, we propose a novel implicit user calibration that focuses on the corneal reflections of binocular eyes. We hypothesize that images reflected on the cornea of binocular eyes extracted around the reflection of the point-of-gaze are similar, and the similarity of images reflected on the cornea is used to identify the kappa angle without constraint, such as screen plane. As a simulation study, we confirmed the hypothesis’s validation and the effectiveness of the proposed method through 3DCG simulation.

Visual defects can be visualized on a clinical chart using ophthalmological instruments to indicate an individual’s perception. This study proposes a novel method to generate a geometrically personalized visualization with blind spots by integrating eye tracking and corneal imaging. The potential of the proposed method to advance visual assistive applications was validated through experiments.

Marker-based gaze estimation has been actively studied for supporting large displays instead of conventional eye-tracking using glints generated by emitting near-infrared light from LEDs. Reference markers can be displayed on a screen as imperceptible markers, and it can be identified using observation. However, the surface of the cornea is not flat; thus, the reflection of the marker is distorted. The distortion of the marker affects the accuracy of the estimation of the corneal center. Therefore, we propose a distorted marker that can be observed as an undistorted marker on the cornea, and the point-of-gaze is estimated using the distorted markers on a model-based approach. The comparative evaluation was performed, and the feasibility of the proposed method was confirmed.

A zoom-tracking system for capturing images of fast-moving flying objects was developed synchronizing a low-latency wide-angle high-speed camera with a zoom camera with a high-speed pan/tilt mechanism. When a target is lost within the angle of the zoom camera, it can be simultaneously tracked again once the target is captured within the angle of view of the wide-angle camera. The effectiveness of this system was demonstrated by shooting a bird drone in real time to tradict with video stabilization even when threre is occlusion with a obstade. And we also confirmed that offline image stabilization can be used for blur-free shooting.

Stereo analysis using multiple cameras is useful for 3D motion analysis of flying objects in the sky. In such stereo analysis, it is necessary to correspond objects detected in different cameras, but it becomes difficult when objects have similar appearances. Motion of flying objects such as wing flapping don’t match perfectly with those of the other flying objects, and it is possible to correspond objects detected in different camera view by checking the sychronicity of their motion. In this study, we propose an algorithm for flying objects that can correspond multiple objects using normalized cross-correlation by extracting motion features of objects in different cameras.

As the issues of declining birthrates, aging population, and urban depopulation become more severe, the agricultural sector is facing an increasingly serious labor shortage. Particularly in the process of harvesting and sorting crops, the lack of sufficient labor has become a significant factor limiting production efficiency. To address this challenge, this paper proposes an automatic identification system based on the YOLO algorithm, specifically targeted at the recognition of radishes in a bin-picking scenario. This system can 　 improve the accuracy and processing speed of radish recognition. Furthermore, this study explores the potential application of this system in actual agricultural production, demonstrating how it can alleviate the labor shortage problem and enhance agricultural production efficiency.

Thermal imaging with long-wavelength infrared (LWIR) cameras is used for various applications including human detection and non-invasive measurements by measuring surface temperature of objects. As a camera that can increase the amount of information using both visible and LWIR lights while avoiding parallax, visible and LWIR coaxial cameras have been built that image both wavelengths on a same optical axis. Among them, a miniaturized coaxial camera that was able to be mounted on mobile robots. However, the camera faced a problem with image quality degradation a few minutes after startup. First, we clarified that the cause of the degradation was due to infrared reflections on the optical window of the camera casing. Second, we redesigned the casing lid with an inclined optical window. Finally, we confirmed that the redesign addressed the issue of infrared reflection.

We propose a method to estimate the position of the opening center of a translucent bag by combining visual and tactile information, which is difficult to estimate accurately only by vision. We verify the proposed method using a 5-DOF arm-type robot equipped with a camera and a tactile sensor on its hand. In the visual estimation, the probability of the opening center of the bag is output as a probability distribution on the image from the input RGB image. In the tactile estimation, the position of the opening center is estimated from the tactile information obtained, and the area where the opening center of the bag cannot exist is estimated. By integrating these estimation results, the position of the opening center is estimated.

Service robot arms often need to change tools depending on the object. However, a universal vacuum gripper (UVG) can be used to suck a variety of objects without changing the tool, enabling stable grasping. In this research, we develop a system to recognize and position the object using 3D data obtained from a depth camera and other sensors and to detect the surface that is easy to grip. The goal is a system that directs the robot arm to the optimum grasping point. This paper describes the results of an attempt to implement 3D object recognition and a plane detection function to estimate readily adsorbed surfaces from point cloud data.

In recent years, research on manipulation of deformable objects has been increased. Our research group has been studying paper manipulation with a multi-fingered hand robot, and has so far succeeded in ”triangular two-folds” with common single-color origami paper. However, more complex folding operations have not been realized. One of the reasons is that the physical simulator used for estimation cannot represent ”folds” of papers. In this study, we proposed a paper shape estimation system using an origami simulator that can represent origami folds, and verified the effectiveness of the proposed system. As a result, the proposed system showed higher tracking performance than the previous system.

Our research group has studied high-speed manipulation by integrating visual information from a large number of cameras installed on a robot body to expand the viewing angle and eliminate blind spots. In this study, a multi-eye vision hand arm is applied to object tracking control. The position of a moving object is calculated from multiple cameras by multi-view stereo, and the robot hand follows it. In the case of multi-view stereo, even if some cameras are obscured by occlusion, stereo calculation is possible with information from other cameras. On the other hand, the effect of calibration errors cannot be ignored, especially for cameras mounted on the robot’s body. Therefore, we verified the performance of multi-view stereo on a multi-eye vision hand arm through experiments.

In our study, we propose a gaze direction Random Access Vision method using a resonant mirror and a multi-tap lock-in pixel image sensor that captures arbitrary and multiple gaze directions at each frame. However, in the conventional method, the timing of gaze direction capturing must be integrated into the system in advance, and only the movement of a predetermined gaze direction is possible. In this study, we propose a hi-speed and dynamic gaze direction selection method that dynamically changes the timing of camera shots during measurement and enables visual feedback such as tracking of the captured object through image processing.

In recent years, research has been conducted on the use of surface electromyography as a method of analyzing arm movements. This technology is currently being used in a wide range of fields, including myoelectric prosthetic hands. Since it uses biological signals emitted by a person during movement, it is possible to realize user-friendly machine operation without the need to learn special operating methods or require large-scale equipment other than sensors. In this study, in order to extend the function of myopotentials as an interface, we used surface myopotentials of proactively moving muscles to identify the movements of reactively moving muscles.

In this research, we create individual hand models with the aim of making simulations in VR space more realistic. In order to create it more easily than traditional methods, we used Blender to process the mesh data created using a 3D scanner into a form that can be hand tracked in Unity. We also conducted hand tracking using Meta Quest2 and Leap motion, and the created individual hand model, and evaluated changes in the visual impression due to the size difference between the created individual hand model and the actual subject’s hand.

Preventing falls and improving gait function in the elderly is crucial for addressing the increasing number of care recipients and extending healthy life expectancy. Elderly individuals with potential for improvement in motor function due to age-related decline in muscle strength are referred to as frail elderly. Therefore, gait measurement experiments were conducted targeting frail elderly individuals. This study focused on a gait parameter called Minimum Foot Clearance (MFC). Decreased MFC increases the risk of stumbling due to contact between the foot and the ground, leading to an increased risk of falls. Therefore, the aim of this study was to elucidate the gait parameters that influence the increase or decrease in MFC. Correlation analysis was conducted using MFC and other gait parameters, revealing that the upper leg angle at MFC was a parameter that may influence changes in MFC.

Environmental recognition by sensors mounted on robots is a pivotal area of research within robotics. This study aims to pioneer a terrain sensing system with multiple thin-film pressure sensors on a spherical mobile robot platform. This system is able to measure the reaction force when the robot makes contact with the ground, thereby obtaining the corresponding stiffness data from different terrain materials. Machine learning techniques are applied to distinguish the terrain characteristics with high accuracy since the data is non-linear, and the patterns of different materials are imperceptible for humans to distinguish.

Masticatory ability is the foundation of our healthy life and it has closely relation with our motor function and cognitive ability. However, measurement of masticatory motion stably is difficult in infancy (0-5 years old) period when basic mastication ability is acquired. in this study, a safe and less invasive lip movement measurement method using an RGB-D (depth) camera has been developed for the analysis of oral functions and feeding behavior. In this report, it is evaluated that the image analysis methods and measured values of feature points by using comparison with 2 kinds of RGB-D cameras and an ultrahigh-speed 3D surface morphology imaging system. As a result, it was confirmed that the constructed system with RGB-D camera D405 has tendency of low error at the specific feature points.

The texture presented by the movement of a large number of specific real objects in a line is called “ SHITSUKAN ”. We can find SHITSUKAN in various places, such as natural objects and arts. This SHITUKAN has been proposed for some time, few studies have evaluated what impression it makes on humans. In this study, we explored how many categories of SHITSUKAN humans classify when they see multiple types of group texture representations. We asked participants to free-sorting categorize 46 image stimuli collected focusing on natural objects based on eight different evaluation words. We present representative figures that were extracted from multidimensional analysis.