論文

© 2019 IEEE. Most SLAM works based on the assumption of static scene, so localization of the camera and mapping of the scene can fail and lose accuracy of the scene, including moving objects. This paper presents a method for simultaneous mapping of moving objects in the target scene and localization of the moving camera, based on geometrical segmentation of each temporal frame. Taking advantage of segmentation of the target scene, using only the geometric structure of the scene, our method can estimate relative pose for camera and every geometrically segmented areas even without recognizing each object. For confirming the effectiveness of the proposed method, we experimentally show that our method can estimate relative poses for all segmented areas in the scene, so that we can achieve SLAM for the scene, including multiple moving objects.

© 2018 The Institute of Electronics, Information and Communication Engineers. We propose a novel object recognition system that is able to (i) work in real-time while reconstructing segmented 3D maps and simultaneously recognize objects in a scene, (ii) manage various kinds of objects, including those with smooth surfaces and those with a large number of categories, utilizing a CNN for feature extraction, and (iii) maintain high accuracy no matter how the camera moves by distributing the viewpoints for each object uniformly and aggregating recognition results from each distributed viewpoint as the same weight. Through experiments, the advantages of our system with respect to current state-of-the-art object recognition approaches are demonstrated on the UW RGB-D Dataset and Scenes and on our own scenes prepared to verify the effectiveness of the Viewpoint-Class-based approach.

© 2018 The Institute of Electronics, Information and Communication Engineers. In this paper, we propose a method to generate a threedimensional (3D) thermal map and RGB + thermal (RGB-T) images of a scene from thermal-infrared and RGB images. The scene images are acquired by moving both a RGB camera and an thermal-infrared camera mounted on a stereo rig. Before capturing the scene with those cameras, we estimate their respective intrinsic parameters and their relative pose. Then, we reconstruct the 3D structures of the scene by using Direct Sparse Odometry (DSO) using the RGB images. In order to superimpose thermal information onto each point generated from DSO, we propose a method for estimating the scale of the point cloud corresponding to the extrinsic parameters between both cameras by matching depth images recovered from the RGB camera and the thermal-infrared camera based on mutual information. We also generate RGB-T images using the 3D structure of the scene and Delaunay triangulation. We do not rely on depth cameras and, therefore, our technique is not limited to scenes within the measurement range of the depth cameras. To demonstrate this technique, we generate 3D thermal maps and RGB-T images for both indoor and outdoor scenes.

Copyright © 2020 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved. Recording surgery is important for sharing various operating techniques. In most surgical rooms, fixed surgical cameras are already installed, but it is almost impossible to capture the surgical field because of occlusion by the surgeon’s head and body. In order to capture the surgical field, we propose the installation of multiple cameras in a surgical lamp system, so that at least one camera can capture the surgical field even when the surgeon’s head and body occlude other cameras. In this paper, we present a method for automatic viewpoint switching from multi-view surgical videos, so that the surgical field can always be recorded. We employ a method for learning-based object detection from videos for automatic evaluation of the surgical field from multi-view surgical videos. In general, frequent camera switching degrades the video quality of view (QoV). Therefore, we apply Dijkstra’s algorithm, widely used in the shortest path problem, as an optimization method for this problem. Our camera scheduling method works so that camera switching is not performed for the minimum frame we specified, and therefore the surgical field observed in the entire video is maximized.

copyright © 2016 Society of Information Display. All rights reserved. We propose a system that supports truck drivers. It can be used to prevent collisions between trucks and obstacles such as humans, bikes, or cars. This is accomplished by using a top view image and a function for the detection of obstacles around a truck based on their height. Our system requires only color cameras.

copyright © 2016 Society of Information Display. All rights reserved. We present a method for synthesizing wide-field-of-view (FOV) RGB-D images by combining three-dimensional (3D) Lidar and an omnidirectional RGB camera. In this system, 3D point clouds captured by the Lidar are upsampled and colored by registration with the omnidirectional RGB image. We show free-viewpoint images generated via this method.

© 2018, Society for Imaging Science and Technology. Vehicle pose estimation is a vital technology for reconstructing the circumstaces of traffic accidents. We propose a novel method for reconstructing the trajectory of vehicles from drive recorder images and a point cloud around the road. First, we apply ORB-SLAM to image sequence of the drive recorder for obtaining the vehicle pose trajectory; however this is based on relative coordinates and a relative scale. For estimating the absolute coordinates and scale of the trajectory, which cannot be obtained from a monocular SLAM like ORB-SLAM, we match the feature points detected in the image sequence with the three-dimensional (3D) point cloud of surrounding environment. For finding 3D points matching the feature points, we generate candidate images by the rendering 3D point cloud of the surrounding environment using the position initially estimated by the Global Positioning System (GPS). Next, we match to obtain the 3D two-dimensional (2D) generated images and drive recorder image to get 3D-2D point correspondences between the 3D point cloud and the drive recorder images; thus, we can convert the relative estimation of the camera pose by ORB-SLAM to the coordinates of the 3D point cloud of the surrounding environment. In the evaluation experiments, we confirmed the effectiveness of our method by comparing the vehicle poses estimated by our method, with those of RTKGPS, which exhibits high measurement precision.

© 2019 Japan Society for Precision Engineering. All rights reserved. The on-the-spot inspections after accident are based on information obtained from the brake marks left on the road and the damage situation of the surroundings. However, this method is affected by the situation before the accident and cannot estimate quantitative guessing. Moreover, technologies such as automatic braking system and lane keeping system, have been developed actively. In developing these technologies, vehicle trajectory is one of the most important data. In this paper, we propose a novel method of estimating vehicle trajectory by using drive recorder images and point cloud which is measured by Mobile Mapping System and LiDAR. In the evaluation experiments, we confirmed the effectiveness of our method by comparing the vehicle poses estimated with RTK-GPS and showing the average errors are within one meter.

© 2019 IEEE. Superquadrics are one of the ideal shape representations for adapting various kinds of primitive shapes with a single equation. This paper revisits the task of representing a 3D human body with multiple superquadrics. As a single superquadric surface can only represent symmetric primitive shapes, we present a method that segments the human body into body parts to estimate their superquadric parameters. Moreover, we propose a novel initial parameter estimation method by using 3D skeleton joints. The results show that superquadric parameters are estimated, which represent human body parts volumetrically.

© Springer Nature Switzerland AG 2019. Superquadrics are one of the ideal shape representations for adapting various kinds of primitive shapes with a single equation. This paper revisits the task of representing a 3D human body with multiple superquadrics. As a single superquadric surface can only represent symmetric primitive shapes, we present a method that segments the human body into body parts to estimate their superquadric parameters. Moreover, we propose a novel initial parameter estimation method by using 3D skeleton joints. The results show that superquadric parameters are estimated, which represent human body parts volumetrically.

© 2017 IEEE. Over 20 years have passed since a free-viewpoint video technology has been proposed in which a user's viewpoint can be freely set up reconstructed 3D space of a target scene photographed by multi-view cameras. This technology allows us to capture and reproduce the real world as it recorded. Once we capture the world in a digital form, we can modify the world as augmented reality does (i.e., placing virtual objects in the digitized real world). As oppose to this concept, the augmented world also allows us to see through real objects by synthesizing the backgrounds that cannot be observed in our raw perspective directly. The key idea is to generate the background image using multi-view cameras observing the backgrounds at different positions, and seamlessly overlaying the recovered image in our digitized perspective. In this article, our studies regarding such desired view generation techniques are reviewed, and its future directions and open problems are discussed.

© 1991-2012 IEEE. This paper proposes a novel algorithm that calibrates multiple cameras scattered across a broad area. The key idea of the proposed method is 'using the position of an omnidirectional camera as a reference point.' The common approach to calibrating multiple cameras assumes that the cameras capture at least some common points. This means calibration becomes quite difficult if there are no shared points in each camera's field of view (FOV). The proposed method uses the position of an omnidirectional camera to determine point correspondence. The position of an omnidirectional camera relative to the calibrated camera is estimated by the theory of epipolar geometry, even if the omnidirectional camera is placed outside the camera's FOV. This property makes our method applicable to multiple cameras scattered across a broad area. Qualitative and quantitative evaluations using synthesized and real data, e.g., a sports field, demonstrate the advantages of the proposed method.

© 2018 Association for Computing Machinery. The first ACM International Workshop on Multimedia Content Analysis in Sports (ACM MMSports'18) is held in Seoul, South Korea on October 26th, 2018 and is co-located with the ACM International Conference on Multimedia 2018 (ACM Multimedia 2018). The goal of this workshop is to bring together researchers and practitioners from academia and industry to address challenges and report progress in mining and content analysis of multimedia/multimodal data in sports. The combination of sports and modern technology offers a novel and intriguing field of research with promising approaches for visual broadcast augmentation, understanding, statistical analysis and evaluation, and sensor fusion. There is a lack of research communities focusing on the fusion of multiple modalities. We are helping to close this research gap with this first workshop of a serious workshops on multimedia content analysis in sports.

© 2018 IEEE. State of the art methods for diminished reality (DR) propagate pixel information from keyframes to later frames to achieve real-time image inpainting in 3D spaces. However, this approach assumes a planar scene and produces artifacts, if the scene geometry is not sufficiently planar. In this paper, we present 3D PixMix, a new real-time inpainting method that addresses non-planar scenes by considering both color and depth information in the inpainting process. We define cost functions for both the color and the geometric appearance in the inpainting scheme and use an RGB-D sensor for depth fusion using a SLAM. Comparison results against the conventional PixMix show that 3D PixMix obtains the equivalent or even better quality in 3D scenes with additional depth information.

© 2018, Society for Imaging Science and Technology. In this paper, we present a refocus interface to set the parameters used for diminished reality (DR)-based work area visualization and a multiview camera-based rendering scheme. The refocus interface allows the user to determine two planes - one for setting a virtual window, through which the user can observe the background occluded by an object, and the other for a background plane, which is used for the subsequent background rendering. The background is rendered considering the geometric and appearance relationships of the multiview cameras observing the scene. Our preliminary results demonstrate that our DR system can visualize the hidden background.

© 2018 Japan Society for Precision Engineering. All rights reserved. This paper presents a method to measure foot shape from a 3D point cloud as input captured from multiple directions using a smartphone depth camera. Such a 3D point cloud could potentially include noise or omit parts of the foot due to occlusion. To deal with this occlusion problem, we propose to use a dataset of 3D foot shapes collected by a precise 3D shape scanner of foot shapes. According to the dataset of 3D foot shapes, we can generate a deformable model by performing a principal component analysis (PCA) on the dataset. Then we minimize the error of the shape represented by the deformable model and the 3D point cloud acquired by the smartphone camera, to recover a complete 3D shape of the entire foot with high accuracy. We test this method by comparing the 3D shape produced by our proposed method to the 3D shape precisely measured by the 3D scanner. Our proposed method can scan the foot shape with an error of about 1,13mm.

© 2019, Springer Nature Switzerland AG. Today, cameras have become smaller and cheaper and can be utilized in various scenes. We took advantage of that to develop a thumb tip wearable device to estimate joint angles of a thumb as measuring human finger postures is important in terms of human-computer interface and to analyze human behavior. The device we developed consists of three small cameras attached at different angles so the cameras can capture the four fingers. We assumed that the appearance of the four fingers would change depending on the joint angles of the thumb. We made a convolutional neural network learn a regression relationship between the joint angles of the thumb and the images taken by the cameras. In this paper, we captured the keypoint positions of the thumb with a USB sensor device and calculated the joint angles to construct a dataset. The root mean squared error of the test data was 6.23° and 4.75° .

Copyright © The Authors, 2018 This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted re-use, distribution, and reproduction in any medium, provided the original work is properly cited.. Over 20 years have passed since a free-viewpoint video technology has been proposed with which a user's viewpoint can be freely set up in a reconstructed three-dimensional space of a target scene photographed by multi-view cameras. This technology allows us to capture and reproduce the real world as recorded. Once we capture the world in a digital form, we can modify it as augmented reality (i.e., placing virtual objects in the digitized real world). Unlike this concept, the augmented world allows us to see through real objects by synthesizing the backgrounds that cannot be observed in our raw perspective directly. The key idea is to generate the background image using multi-view cameras, observing the backgrounds at different positions and seamlessly overlaying the recovered image in our digitized perspective. In this paper, we review such desired view-generation techniques from the perspective of free-view point image generation and discuss challenges and open problems through a case study of our implementations.

©2019 Association for Computing Machinery. This paper presents a 3D virtual eyeglasses try-on system for practical use. For fitting eyeglasses in a shop, consumers wish to look at themselves in a mirror while trying on various eyeglass styles. However, for people who need to wear eyeglasses for correcting problems with eyesight, it is impossible for them to clearly observe their face in the mirror without wearing eyeglasses. This makes fitting them for new eyeglasses difficult. This research proposes a virtual try-on system that can be used while wearing eyeglasses.We replace the user's eyeglasses in the input video with new eyeglasses virtually. Moreover, a fast and accurate face tracking tool enables our system to automatically display 3D virtual glasses following a user's head motion. Experimental results demonstrate that the proposed method can render virtual glasses naturally while the user is wearing real eyeglasses.

© 2019 IEEE. Recording medical surgery operations is important for sharing the various operating techniques. In most operating rooms, fixed surgery cameras are already installed, but it is almost impossible to capture the surgical field because of occlusion by the surgeon's head and body. In order to capture the surgical field in real surgery operations, we proposed the installation of multiple cameras in a surgical lighting system, so that at least one camera can capture the target surgical field even when the surgeon's head and body occlude other cameras. In this paper, we present a method for automatic viewpoint selection from multi-view surgical videos, so that the surgical field can always be recorded in the output video. We employ a method for learning-based object detection from videos for automatic evaluation of the surgical field area from multiple input videos. By selecting the viewpoint with the largest area of the surgical field, we can virtually reduce the area of the surgeon's head and hands. In general, frequent camera switching degrades the video quality of view (QoV). Therefore, we apply the Dijkstra method widely used in the shortest path problem as a combinatorial optimization method for this problem. Our camera scheduling method is that the camera switching is not performed for a certain period of time, and the surgical field observed in the entire video is maximized.