88
(3)物体の全周を計測して得られた形状データから,物体の位相に関する事前知識をもっ ことなく任意の位相に適応して柔軟に全周型物体モデノレを生成する仕組みを実現した.
本手法では,等高面法と最適化手法を用いることにより,これまでモデリングの対象が 球面と同相な物体に限られていたのに対し,穴領域を有する物体や複数の物体からな
る三次元シーンなどに対するモデリングも可能になった.
謝 辞
本論文は,筆者が福岡県工業技術センター機械電子研究所職員として在職中および九州大 学大学院システム情報科学研究科知能システム学専攻博士後期課程在学中に行なった研究を まとめたものです.
本研究科において,知能システム学専攻長谷川勉教授には,本論文のまとめ方など終始親 身に多大な御指導,御助言および御鞭捷を賜わりました.ここに心からの謝意を表します.ま た,査紅彬助教授には,本研究を遂行するうえで,終始親身に多大な御指導,御助言および 御鞭掻を賜わりました.ここに心から感謝いたします.さらに,本論文をまとめるにあたり,
本研究科知能システム学専攻迫江博昭教授,同知能システム学専攻谷口倫一郎教授より有益 な御批評を頂きました.ここに心から感謝する次第です.
工学部情報工学科長田正教授(現九州システム情報技術研究所所長)には,終始親身に多 大な御指導,御助言および御鞭捷を頂きました.ここに心から感謝する次第です.
木室義彦講師(現九州、│システム情報技術研究所研究員),岡田伸鹿講師には,有益な御教示 と御助言を頂きました.お二人に心から感謝いたします.桐木利弘助手,江藤淳二技官には 有益な御助言と御協力を頂きました.お二人に心から感謝いたします.
九州大学大学院システム情報科学研究科長谷川研究室および査研究室の大学院生,学部生 および卒業生の皆様には熱心な御討論を頂きました.ここに深く感謝いたします.
松永純一氏(現飯塚研究開発機構研究開発部長)をはじめ,福岡県工業技術センター機械 電子研究所の先輩諸氏,同僚の皆様には,多くの御助言と常に暖かな励ましをいただきまし た.ここに深く感謝いたします.
最後に,博士課程に進学し,研究生活を送ることに対して深い理解を示し,応援していた だいた妻徳子と両親に心から感謝いたします.
1999年9月
参 考 文 献
[1] L. G. Ro berts. Machine perception of three‑dimensional solids
,
optical and electro‑ optical information processing. Technical report,
MIT Press,
1965.[2] A. Guzman. Decomposition of a visual scene into three‑dimensional bodies. In Proc. of Fall Joint Comp. Con
, . f
1968.[3] M. Oshima and Y. Shirai. Object recognition using three‑dimensional information. 1EEE Trans. PAλ11
,
Vo1.5,
No, 4 .
pp.353・361,
1983.[4] P. Besl and R. Jain. Invariant surface characteristics for 3D object recog凶tionin range data. Computer・VisionGr・αphicsαnd1mαge Processing
,
Vo1.33,
pp.33‑80,
1986. [5] T. J. Fan,
G. Medio∞
n爪1tion. 1EEE Trαηs. PAM1
,
Vo.111,
No.11,
pp.1140・1157,
1989.[6] P. J. Besl and N. D. McKay. A method for registration of 3・Dshapes. 1EEE Trans. PAλ11
,
Vol.14,
No.2,
pp.239・256,
1992.[7] Y. Chen and G. Medioni. Object modelling by registration of multiple range images. 1mαgeαnd Vision Computing, Vol.10
,
pp.145・155,
1992.[8] H. Zha
,
K. Morooka,
and T. Hasegawa. Active modeling of 3・Dobjects: planning on the next best pose (NBP) for acquiring range images. In Proc. of 1nt. Conf. on Recent Advances in 3‑D Digital 1mαging and Modeling,
pp.68‑75,
1997.[9] T. Poggio
,
V. Torre,
and C. Koch. Computational vision and regularization theory. Nature, Vo.1317,
pp.314‑319,
1985.[10] A. N. Tikhonov and V. Y. Arsenin. Solutions of Ill‑Posed Problems. Winston and Wiley Publishers. Washington D.C.
,
1977.[11] S. Geman and D. Geman. Stochastic relaxation
,
Gibbs distributions and the Bayesianrestoration of images. 1EEE Trαns. PAM1
,
Vo1.6,
No.6,
pp.721・741,
1984.参 考 文 献 91
[12] J. L. Marroquin. Deterministic Bayesian estimation of Markovian random fields with applications to computational vision. In Proc. of Int. Conf. on Computer Vision
,
1987.
[13] E. B. Gamble and T. Poggio. Visual integration and detection of discontinuities : The key role of intensity edges. A r・iificialIntell. Lαb.
,
Massαchusetts Inst. Technol.,
AI Memo
,
No.970ヲ1987.[14] T. Poggio
,
E. B. Gamble,
and J. J. Little. Parallel i凶nt旬egra州1Science
,
Vo.1242,
pp.436・440,
1988.[15] S. Z. Li. Markov Random Field Modeling in Computer Vision. Berlin:Springer‑Verlag
,
1995.
[16] S.‑Z. Li. Towards 3D vision from range images: An optimization framework and parallel networks. Computer・VisionGraphics and Image Processing
,
Vo1.24,
pp.52・96
,
1983.[17]浦浜喜一,渡辺 真.画像の非線形尺度空間フィノレタリング:連続変形法と非等方拡散.
電子情報通信学会論文誌, Vol.J77・D‑II,No.9, pp.1761‑1767, 1994.
[18] V. Granville
,
M. Krivanek,
and J. P. Rasson. Simulated annealing: A proof of con‑ vergence. IEEE 7干αns.PAMI,
Vol.16,
pp.652・656,
1994.[19] D. Geiger and F. Girosi. Parallel and deterministic algorithms from MRF's: Surface reconstruction. IEEE Trαηs. PAMI
,
Vol.13, No.5, pp.401・412,1991.[20] G. Whitten. A framework for adaptive scale space tracking solutions to problems in computational vision. In Proc. of Int. Conf. on Computer・Vision
,
pp.210‑220,
1990. [21 ]横矢直和.多重スケールでの正則化によるステレオ画像からの不連続を保存した曲面再構成.電子情報通信学会論文誌, Vo.lJ76・D‑II,No.8, pp.1667‑1675, 1993.
[22] P. Craven and G. Wahba. Smoothing noisy data with spline functions ‑Estimating the correct degree of smoothing by the method of generalized cross‑validation. Numerische Mαthematik
,
No.31,
pp.377‑403,
1979.[23] B. Shahraray and D. J. Anderson. Optimal estimation of contour properties by cross‑ validated regularization. IEEE Trans. PAMI
,
Vo1.11, No.6, pp.600・610,1989.参 考 文 献 92
[24] G. Wahba and J. Wendelberger. Some new mathematical methods for variational objective analysis using splines and cross‑validation. Monthly Weather Rev.
,
Vol.108,
pp.36・57
,
1980.[25] L. D. Cohen and
. r
Cohen. Finite‑element methods for active contour models and balloons for 2・D and 3‑D images. 1EEE 1トαns.PAM1,
Vol.15,
No.11,
pp.1131・1147,
1993.
[26] D. Terzopoulos and M. Vasilescu. Sampling and reconstruction with adaptive meshes. In Proc. of Computer Vision αnd Pattern Recognition
,
pp.70‑75,
1991.[27] H. Delingette
,
M. Hebert,
and K. Ikeuchi. 8hape representation and image segmenta・tion using deformable surfaces. In Proc. of Compute1・Visionαnd Pαttern Recognition
,
pp
. 4
67・472,
1991.[28] S. Chen
,
G. Stockman,
and K. Chang. 80 dynamic deformation for building of 3・D models. 1EEE 1干αηs.N N,
Vo1.7,
pp.374・387,
1996.[29] K. Higuchi
,
M. Hebert,
and K. Ikeuchi. Building 3・Dmodels from unregistered range images. In Proc. of 1nt. ConJ .
on Robotics and Automation,
pp.2248・2253,
1994. [30] T. McInerney and D. Terzopoulos. A finete element model for 3D shape reconstructionand nonrigid motion tracking. In Proc. of 1nt. Conf. 0η Compute1・Vision
,
pp.518・523,
1993.
[31] A. J. Bulpitt and N. D. Efford. An efficient 3D deformable model with a self‑optimising mesh.1mαge and Vision Computing
,
Vo.l14,
pp.573・580,
1996.[32] H. Hoppe
,
T. DeRo民 T.Duchamp,
J. McDo∞
nation. In P1・oc.of S1GGRAPH 93
,
Vo1.27,
pp.19・26,
1993.[33] M. Kass
,
A. Witkin,
and D. Terzopoulos. Snakes: Active .contour models. 1nt. J. 01 Compute1・Vision,
Vo1.1,
No, 4 .
pp.321‑331,
1987.[34] D. Terzopoulos
,
A. Witkin,
and M. Kass. Constraints on deformable models: Recov‑ ering 3D shape and nonrigid motion. A 1・tificial1ntelligence,
Vo.136,
No.1,
pp.91・123,
1988.
[35] R. Malladi
,
J. A. Sethian,
and B. C. Vcm凶 .Shape modeling with front propagation: A level set approach. 1EEE Trαηs. PAM1,
Vo.l17,
No.2,
pp.158・175,
1995.参 考 文 献 93
[36] V. Caselles
,
R. Kimmel,
and G. Sapiro. Geodesic active contours. In Proc. of 1nt. Conf. on Compute1・Vision,
pp.69ι699,
1995.[37] A. Blake and A. Zisserman. Visual Reconstruction
,
Cambridge,
MA:MIT Press,
1987. [38] F. Spitzer. Markov Random Fields and Gibbs Ensembles. Amer.MIαth.Mon, ・
Vo1.78,
pp.142‑154
,
1971.[39] D. Terzopo山 sand D. Metaxas. Dynamic 3D models with local and global deforma‑ tions: Deformable superquadrics. 1EEE T:同ns.PAM1
,
Vol.13,
pp.703‑714,
1991. [40] D. Metaxas and D. Terzopoulos. Shape and nongirid motion estimation throughphysics‑based synthesis. 1EEE 1干αns.PAλ11, Vol.15, pp.580・591,1993.
[41] D. Metax民 E.Koh
,
and N. 1. Badler. Multi‑level shape representation using global deformation and locally adaptive finite elements. Int. J. of Computer Vision,
Vo1.25,
No.1
,
pp.49・61,
1997.[42] J. Park
,
D. Metaxas,
and A. Young. Deformable models with parameter functions: Application to heart‑wall modeling. In Proc.0 1
Computer Vision αnd Pattern Recog‑ nition,
pp.437‑442,
1994.[43] D. Terzopoulos. Reg山 .rizationof inverse visual problems involving discontinuities. 1EEE Trαηs. PAλ11, Vo1.8, No, .4 pp.413・424,1986.
[44] G. Dhatt and G. Touzot. The Finite Element Method Displayed. Wiley
,
New York,
1984.
[45] C. Rao. Adv α nced statistical methods i灼η
胸
biωomηnetr巾 T陀reωsseaα灯M7ア 川 .[46]長尾 真監訳.ディジタノレ画像処理.近代科学社, 1972.
[47] P. W. M. Tsang
,
P. C. Yuen,
and F. K. Lam. Recognition of occIuded objects. Pattern Recognition,
Vo1.25,
pp.1107‑1117,
1992.[48] D. Terzopoulos. The computation of visible‑surface representations. 1EEE Trans. PAM1
,
Vo.110, No.4, pp.417・438,1988.[49] T. E. Boult and A. D. Gross. Recovery 0ぱfs叩e即
Proc. of SP1E 1ntell. Robots αnd Comp. Vision
,
pp.358・365,
1987.参 考 文 献
94[50] Y. Sato
,
K. Hattori,
and M. Ot叫<I.Real‑time handy rangefinder Cubicscope. In Proc. of Int. Conf. on Automation,
Robotics αnd Computer・Vision,
pp.1148・1152,
1994. [51] S. Kumar and D. Goldgof. Model based part segmentation of range data: 'hyper‑quadrics and dividing planes. In Pr・oc.of Wor・kshopPhysics‑Bαsed M odeling in Com‑
puter Vision
,
pp.17・23,
1995.[52] H. Zha, T. Hoshide, and T. Hasegawa, A recursive日ttingand splitting algorithm for 3・
D object modeling using superquadrics. In Proc. of Int. Conf. on Pattern Recognition
,
pp.658・662
,
1998.[53] M. R. Korn and C. R. Dyer. 3・D multiview object representations for model‑based object recognition. Pαttern Recognition.
,
Vo1.20, No.1, pp.91‑103, 1987.[54]坂根茂幸,石井 優,柿倉正義.ハンドアイ行動シミュレーク :'HEAVENシステムに基 づく視覚センサのオクルージョン回避.日本ロボット学会誌.Vo1.4, No.4, pp.382・392, 1986.
[55J D. J. Williams and M. Shah. A fast algorithm for active contours. In Proc. of Int. Con
f .
on Computer Vision,
pp.592・595,
1990.[56] C. Koch
,
J. Marroquin,
and A. Yuille. Analog neuronal networks in early vision. In Pr・oc.of Natl. Acαd. Sci. USA,
pp.4263‑4267,
1986.[57]久長穣,山下 雅史,阿江忠.ホップフィールドニューラルネットワークによる実数 集合の分割.電子情報通信学会論文誌, Vol.J73‑D‑I ,I No.8, pp.1213・1219,1990. [58] S. Osher and J. A. Sethian. Fronts propagating with curvature dependent
speed:Algorithms based on Hamilton‑Jacobi formulation. J. of Computationαl Physics
,
Vol. 79,
pp.12・49,
1988.[59] J. A. Set1山n.Le似 SetMethods. Cambridge University Press
,
1996.[60] J. A. Sethian and J. Strain. Crystal growth and dendritic solidification.よ ofCom‑
putationα1 Physics
,
Vo1.98,
pp.231・253,
1992.[61 ]儀我 美ーほか.動く曲面を追いかけて.日本評論社, 1996.
[62] D. L. Chopp. Computing minimal surfaces via level set curvature flow. J. of Comp1ト
αttionα1 Physics
,
Vo.1106,
pp.77‑91,
1993.参 考 文 献 95
[63] R. T. Whitaker. A level‑set approach to 3D reconstruction from range data. Int.よ of Compute1・Vision