3,a) 3 3 ( ) DB 3D DB 2D,,,, PnP(Perspective n-point), Ransac. DB [] [2] 3 DB Web Web DB Web NTT NTT Media Intelligence Laboratories, - Hikarinooka Yokosuka-Shi, Kanagawa 239-0847 Japan a) yabushita.hiroko@lab.ntt.co.jp (A) (B) クエリ画像 学習画像群 クエリ映像 学習画像 仮想視点画像群生成 3 次元推定 仮想視点群統合画像生成 マッチング マッチング 認識結果 認識結果 3 A B 3 A [3][4] 2 B [5][6] 3 2 DB c 202 Information Processing Society of Japan
2 DB DB DB 3 P 2D P 3D 3 [7] DB [6] DB DB 2D [8][9][0] [][2] DB 3 2 3 3 4 2. 2. (B) 2 3 [3] 3 2.2 3 3 c 202 Information Processing Society of Japan 2
2 3 4 5 6 6 2 3 4 5 0 9 7 8 9 0 8 7 3 3 3 3 r 3 2 3 (x orig., y orig., z orig. ) ϑ, φ 3 (x, y, z) (x, y, z) = r (x orig., y orig., z orig. ),() x 2 orig. + y2 orig. + z2 orig. r x2 + y ϑ = 2 + z 2 ϑ x,y,z φ φ x,y ϑ x,y,z, φ x,y cos ϑ x,y,z = cos φ x,y = z x2 +y 2 +z 2 x, sin φ x 2 +y 2 x,y =. y x 2 +y 2 (2) (0 ϑ x,y,z π, 0 φ x,y 2π). (3) 3 (x, y, z) 2 (X, Y ) { X = ϑ Y = φ. (4) 2 (X, Y ) 3 3 (x orig., y orig., z orig. ) 2.3 DB 2.2 3 3 DB 2 4 6 7 0 3 3 DB 2.3. 3D 2D 3 2 PnP Perspective-n point [4][5][6] RANSAC Random Sample Consensus 3 DB 2 (i) DB (ii) (iii) (ii) 6 (iv) (iii) 3 3 DB (v) (iii) (iv) 2.3.2 2.3. 4 n n > (a) 3 2.3. DB (b) DB 2 c 202 Information Processing Society of Japan 3
DB (c) i S i DB 2 d i D i offset offset 2 Score(S i ) = (5) d i + offset D i + offset 2, (A) (B) (d) (b) (c) DB DB DB 3. 3 3 3 KIT ObjectModels Web Database[7] 3 HP DB 3 = 5 A B 3. 2.3 5 5 6 6 7 7 3.2 2 2 DB [6] 8, 9 8 8 DB c 202 Information Processing Society of Japan 4
(a) (I) (b) (II) 6 a b 7 I II 50 I II DB 9 DB 4. 3 3 ( ) DB 3D DB 2D DB DB 20 48 Web DB 3 3 c 202 Information Processing Society of Japan 5
8 [6] DB 200, vol.4, pp.250-26, Nov.200. [0] J. Kim, and K. Grauman, Asymmetric Region-to-Image Matching for Comparing Images with Generic Object Categories, CVPR200, pp.2344-235. [] M. Cho, and K. M. Lee, Progressive Graph Matching: Making a Move of Graphs via Probabilistic Voting, Proc. CVPR202, Jun.202. [2],,,, 3, FIT20, 3, H-03, pp., Sept.20. [3] J. Shimamura, M. Morimoto, and H. Koike. Robust vslam for dynamic scenes. MVA20, pp.344-347, Jun.20. [4], PnP, 90 90, 4-50, 990. [5] F. Moreno-Noguer, V. Lepetit, and P. Fua, EPnP: Accurate Non-Iterative O(n) Solution to the PnP Problem., International Journal of Computer Vision archive, vol.8(2), pp.55-66, Feb.2009. [6],,,, MIRU2005, 2005 7. [7] KIT ObjectModels Web Database, http://i6p09.itec.uni-karlsruhe.de/objectmodelswebui. 9 DB [],, 2 3 -, (D-II), vol.j77-d-ii, no., pp.279-287, Nov.994. [2],,,, Analysis- Synthesis 3 GPU, SSII200, IS4-7, Jun.200. [3] S. Savarese and L. Fei-Fei. View synthesis for recognizing unseen poses of object classes. ECCV2008, vol.5304/2008, pp.602-65, Oct.2008. [4] H. Su, M. Sun, L. Fei-Fei, and S. Savarese. Learning a dense multi-view representation for detection, viewpoint classification and synthesis of object categories., ICCV2009, pp.23-220, Sept.2009. [5],,,, 3,, p.56, 20 3. [6],,, 3,, vol., no. 379, PRMU20-53, pp. 73-78, 202. [7] O. Boiman, E. Shechtman, and M. Irani. In defence of nearest-neighbor based image classification. CVPR2008, pp.-8, Jun.2008. [8],,,, Voting, 40(2), 34-323, 20 3. [9] M. Klinkigt, and K. Kise, From Local Features to Global Shape Constraints: Heterogeneous Matching Scheme for Recognizing Objects under Serious Background Clutter, Asian Conference on Computer Vision c 202 Information Processing Society of Japan 6