Misc - OGAWARA Koichi

動的計画法を用いた多指ハンドによる物体操作運動の生成とフィードバック制御による運動補償

中川 圭祐, 小川原 光一

第32回ロボット学会学術講演会 予稿集     1 - 4   2014.09

ロボットによるSFM法を用いた未知小物体の全周3次元形状推定

古川 直人, 小川原 光一

第19回知能メカトロニクスワークショップ 予稿集     1 - 5   2014.07

動的計画法とフィードバック制御を用いた多指ハンドによる持ち替えを伴う物体操作の獲得

中川 圭祐, 小川原 光一

第19回知能メカトロニクスワークショップ 予稿集     1 - 4   2014.07

画像処理に基づく細胞内構造体の配置の規則性に関する定量解析

小川原 光一

バイオイメージ・インフォマティクス ワークショップ 2014     1 - 1   2014.06

Non-rigid Registration with Reliable Distance Field for Dynamic Shape Completion

Kent Fujiwara, Hiroshi Kawasaki, Ryusuke Sagawa, Koichi Ogawara, Katsushi Ikeuchi

Workshop on Dynamic Shape Measurement and Analysis 2014 in conjunction with International Conference on 3DV     1 - 8   2014  [Refereed]

Gait identification robust to changes in walking direction by 4D gait database

BABA Ryosuke, IWASHITA Yumi, OGAWARA Koichi, KURAZUME Ryo

IEICE technical report. Speech ( The Institute of Electronics, Information and Communication Engineers )  111 ( 431 ) 59 - 64   2012.02

　View Summary

In person identification using gait images, various inherent image features of individuals are extracted from a sequence of gait images. However, for instance, if a subject's observation angle changes compared with those in the database, the correct classification rate gets low. To deal with this problem, we constructed a 4D gait database consisting of multiple 3D shape models of walking subjects, and introduce a method robustly against walking direction changes. In this method, firstly we reconstruct 3D models of subjects from gait images taken by multiple cameras, and then synthesize virtual images of 3D models from multiple arbitrary virtual viewpoints and build a database from gait features extracted from virtual images. In the identification phase, the person is identified by matching the gait features of the subject and those from all virtual viewpoints in the database. However, the calculation cost is expensive due to full search, and the subject is wrongly estimated due to wrong estimation of walking direction. So in this paper, to achieve the reduction of calculation time and high correct classification rate, we introduce a method which estimates the walking direction using Frieze Patterns firstly and then identify the person using features from the estimated virtual viewpoint. Experiments using the 4D gait database show the effectiveness of the proposed method.

Gait identification robust to changes in observation angle

IWASHITA Yumi, BABA Ryosuke, OGAWARA Koichi, KURAZUME Ryo

電子情報通信学会技術研究報告. SP, 音声   111 ( 431 ) 65 - 70   2012.02

Sequential Pattern Recognition by Local Classifiers and Dynamic Time

福冨 正弘, 小川原 光一, フォン ヤオカイ

情報処理学会研究報告 ( 情報処理学会 )  2010 ( 3 ) 8p   2010.10

Sequential Pattern Recognition by Local Classifiers and Dynamic Time Warping

FUKUTOMI Masahiro, OGAWARA Koichi, FENG Yaokai, UCHIDA Seiichi

Technical report of IEICE. PRMU ( 一般社団法人電子情報通信学会 )  110 ( 187 ) 157 - 164   2010.08

　View Summary

This paper describes a method for recognizing sequential patterns with nonlinear time warping. The proposed method uses a sequence of local classifiers, each of which is prepared to provide a recognition result (i.e., class label) at a certain sample point. In addition, in order to compensate nonlinear time warping, the local classifier of the point v has to be assigned to the point t_v of the prototype sequential pattern. Consequently, we must solve the optimal labeling problem and the optimal point-to-point correspondence problem (i.e., the optimal mapping from v to t_v) simultaneously. In the proposed method, this multiple optimization problem is tackled by graph cut. Specifically, the α-expansion algorithm, which is an approximation algorithm for graph cut problems, is employed. After the solving the problem, the input pattern is recognized based on majority voting of the class labels obtained at the local classifiers. Several penalties are introduced for forcing neighboring local classifiers to have the same class labels and continuous point-to-point correspondence. For observing the validity of the proposed method, it was applied to an online character recognition task.

Articulated structure from motion using subspace separation

松尾 幸治, 小川原 光一, 倉爪 亮

IEICE technical report ( 電子情報通信学会 )  109 ( 376 ) 273 - 278   2010.01

部分的に局所性を保持するハッシュ関数を用いた画像列からの高頻度パターン検出

小川原光一, 田邉康史, 倉爪亮, 長谷川勉

日本ロボット学会学術講演会予稿集(CD-ROM)   28th   2010

部分的に局所性の高いハッシュ関数を用いた頻出運動パターンの効率的な抽出法

小川原光一, 田邉康史, 倉爪亮, 長谷川勉

ロボティクスシンポジア予稿集   15th   2010

Detecting Frequent Motion Patterns using Motion Density

田邉康史, 小川原光一, 倉爪亮, 長谷川勉

日本ロボット学会学術講演会予稿集(CD-ROM)   27th   2009

Estimation Method of the pose of an Object Manipulated by a Multi-fingered Robot-hand using Visual and Force Information

亀崎康之, 小川原光一, 倉爪亮, 長谷川勉

日本ロボット学会学術講演会予稿集(CD-ROM)   27th   2009

Manipulation Recognition based on relative motions between two objects using Dynamic Programming

Yonemasu Yoshihiro, Uchida Seiichi, Ogawara Koichi

Record of Joint Conference of Electrical and Electronics Engineers in Kyushu ( 電気・情報関係学会九州支部連合大会委員会 )  2009 ( 0 ) 435 - 435   2009

　View Summary

本論文では，物体操作作業を２物体間の時系列相対運動パターンとみなし，これをあらかじめ用意した参照相対運動パターン集合と比較することによって，動画像から物体操作を認識する手法を提案する．ある時刻の画像に対し，画像中の任意の２領域がある参照相対運動パターンに属する尤度を，(1)１つ目の領域が１番目の物体である尤度，(2)２つ目の領域が２番目の物体である尤度，(3)領域間の相対位置，の３つの指標により定義する．本手法では，この尤度に基づき動的計画法を用いて入力動画像全体と各参照相対運動パターンとの比較を行い，最も類似した参照相対運動パターンが表す物体操作を認識結果として出力する．

DOI

Fast Repetitious Motion Pattern Detection Method via Dynamic Programming using Motion Density

Koichi Ogawara, Yasufumi Tanabe, Ryo Kurazume, Tsutomu Hasegawa

IEEE International Conference on Robotics and Automation     2009

DOI

Obtaining Manipulation Skills from Observation

Takamatsu Jun, Tominaga Hirohisa, Ogawara Koichi, Kimura Hiroshi, Ikeuchi Katsushi

Monthly journal of the Institute of Industrial Science, University of Tokyo ( The University of Tokyo )  52 ( 5 ) 231 - 236   2000.05

DOI