A5071725508- Robotics and Sensor-Based Localization
- Robotic Path Planning Algorithms
- Robotics and Automated Systems
- Video Surveillance and Tracking Methods
- Advanced Vision and Imaging
- Advanced Image and Video Retrieval Techniques
- Autonomous Vehicle Technology and Safety
- Robot Manipulation and Learning
- Advanced Neural Network Applications
- Human Pose and Action Recognition
- Smart Agriculture and AI
- Modular Robots and Swarm Intelligence
- Remote Sensing and LiDAR Applications
- IoT-based Smart Home Systems
- Indoor and Outdoor Localization Technologies
- AI-based Problem Solving and Planning
- Social Robot Interaction and HRI
- Reinforcement Learning in Robotics
- Image and Object Detection Techniques
- 3D Surveying and Cultural Heritage
- Anomaly Detection Techniques and Applications
- Gait Recognition and Analysis
- Hand Gesture Recognition Systems
- Face and Expression Recognition
- Context-Aware Activity Recognition Systems
Toyohashi University of Technology
2015-2024
Okayama University
2023
Tokyo University of Agriculture and Technology
2023
Hokkaido University of Science
2022
The University of Tokyo
1988-2020
Japan Research Institute
2020
Osaka University
2000-2018
Shibuya University Network
2010
Hitachi (Japan)
1998-2009
Craft Engineering Associates (United States)
2005
It is important to measure and analyze people behavior design systems which interact with people. This article describes a portable measurement system using three-dimensional LIDAR. In this system, an observer carries the equipped Light Detection Ranging (LIDAR) follows persons be measured while keeping them in sensor view. The estimates pose environmental map tracks target persons. enables long-term wide-area measurements are hard for existing tracking systems. As field test, we recorded of...
The scene rigidity is a strong assumption in typical visual Simultaneous Localization and Mapping (vSLAM) algorithms. Such limits the usage of most vSLAM dynamic real-world environments, which are target several relevant applications such as augmented reality, semantic mapping, unmanned autonomous vehicles, service robotics. Many solutions proposed that use different kinds segmentation methods (e.g., Mask R-CNN, SegNet) to detect objects remove outliers. However, far we know, kind wait for...
This paper presents an active vision system for real-time traffic sign recognition. The is composed of two cameras, one equipped with a wide-angle lens and the other telephoto-lens, PC image processing board. first detects candidates signs in using color; intensity, shape information. For each candidate, telephoto-camera directed to its predicted position capture candidate larger size image. recognition algorithm designed by intensively built-in-functions off-the-shelf board realize both...
Pedestrian detection is one of the key technologies for autonomous driving systems and assistance systems. To predict possibility a future collision, these have to accurately recognize pedestrians as far away possible. Moreover, function detect not only people walking but also who are standing near road required. This paper proposes method recognizing by using high-definition LIDAR. Two novel features introduced improve classification performance. One slice feature, which represents profile...
Host-associated microbiota is often acquired by horizontal transmission of microbes present in the environment. It hypothesized that differences environmental pool colonizers can influence community assembly on host and as such affect holobiont composition fitness. To investigate this hypothesis, host-associated invertebrate eco(toxico)logical model Daphnia was experimentally disturbed using different concentrations antibiotic oxytetracycline. The host-microbiota interactions when were...
Visual simultaneous localization and mapping (vSLAM) are considered a fundamental technology for augmented reality intelligent mobile robots. However, rigid scene assumption is common in vSLAM, which limits the wide usage populated real-world environments. Recently, with widespread use of artificial neural networks, many solutions have tried to eliminate influence dynamic objects using semantic information provided by object detection or segmentation. Mask R-CNN popular applications, but...
The paper proposes a method to precisely estimate the pose (joint angles) of moving human hand and also refine 3D shape (widths lengths) given model from monocular image sequence which contains no depth data. First, an initial rough shaped model, possible candidates are generated in search space efficiently reduced using silhouette features motion prediction. Then, selecting with high posterior probabilities, poses obtained feature correspondence is resolved even under quick self occlusion....
Autonomous driving systems need to handle complex scenarios such as lane following, avoiding collisions, taking turns, and responding traffic signals. In recent years, approaches based on end-to-end behavioral cloning have demonstrated remarkable performance in point-to-point navigational scenarios, using a realistic simulator standard benchmarks. Offline imitation learning is readily available, it does not require expensive hand annotation or interaction with the target environment, but...
Focusing on the task of point-to-point navigation for an autonomous driving vehicle, we propose a novel deep learning model trained with end-to-end and multi-task manners to perform both perception control tasks simultaneously. The is used drive ego vehicle safely by following sequence routes defined global planner. part encode high-dimensional observation data provided RGBD camera while performing semantic segmentation, depth cloud (SDC) mapping, traffic light state stop sign prediction....
This paper describes an interactive vision system for a robot that finds object specified by user and brings it to the user. The first registers models automatically. When specifies object, tries recognize recognition result is shown user, may provide additional information via speech such as pointing out mistakes, choosing correct from multiple candidates, or giving relative position of object. Based on advice, again Experiments are described using real-world refrigerator scenes.
In soccer games, understanding the movement of players and ball is essential for analysis matches or tactics. this paper, we present a system to track estimate their positions from video images. Our tracks by extracting shirt pants regions can cope with posture change occlusion considering colors, positions, velocities in image. The extracts candidates using color motion information, determines among them based on continuity. To determine player who holding ball, position field 3D are...
This paper describes a map generation method using an omnidirectional stereo and laser range finder. Omnidirectional has advantage of 3D acquisition, while it may suffer from low reliability accuracy in data. Laser finders have reliable acquisition data, they usually obtain only 2D information. By integrating these two sensors, can be generated. Since the sensors detect different parts object, separate probabilistic grid is first generated by temporal integration data each sensor The...
To recognize and retrieve soccer game scenes, the movement of players ball must be analyzed. This paper describes a method tracking estimating their 3D positions from video images recorded TV broadcasting. Our system detects tracks them by predicting motions. determines among candidates considering motion continuity ball. The camera parameters (Pan, tilt, zoom) are estimated defecting line regions on field in every frame matching to model. Because our uses not only straight lines but also...
We have developed a new type of robot which accompanies the healthcare professionals making medical rounds in patient's bedrooms at hospitals. This novel mainly executes two tasks: carrying armamentarium round supplies and recording electronic health data on rounds. An omni-directional mobile mechanism human tracking control system to follow specified realize smooth transfer movement from nurses' station bedroom. Electronic is automatically recorded by using CCD camera voice recorder. When...
This letter presents an interactive graph SLAM framework with a 3D LIDAR. allows the user to interactively correct environmental map generated by automatic system. By optimizing pose consisting of constraints created and correction constraints, which are through graphical interface, we obtain large globally consistent map. We propose semi-automatic loop closing plane-based techniques for creating constraints. also devise constraint update approach refine given SLAM. The evaluation results...
We present a novel compact deep multi-task learning model to handle various autonomous driving perception tasks in one forward pass. The performs multiple views of semantic segmentation, depth estimation, light detection and ranging (LiDAR) bird's eye view projection simultaneously without being supported by other models. also provide an adaptive loss weighting algorithm tackle the imbalanced issue that occurred due plenty given tasks. Through data pre-processing intermediate sensor fusion...