IPSJ SIG Technical Report Vol.2010-CVIM-170 No /1/ Visual Recognition of Wire Harnesses for Automated Wiring Masaki Yoneda, 1 Ta

Size: px

Start display at page:

Download "IPSJ SIG Technical Report Vol.2010-CVIM-170 No /1/ Visual Recognition of Wire Harnesses for Automated Wiring Masaki Yoneda, 1 Ta"

ありさかせ
4 years ago
Views:

1 Visual Recognition of Wire Harnesses for Automated Wiring Masaki Yoneda, 1 Takayuki Okatani 1 and Koichiro Deguchi 1 This paper presents a method for recognizing the pose of a wire harness in a given image; wire (or cable) harnesses are commonly used in the wiring of automobiles. Wire harnesses can be categorized into a difficult class of objects for visual recognition because of the high degrees of freedom of its pose variation due to its flexibility as well as lack of rich image features. To cope with these difficulties, the proposed method represents a wire harness by a combination of the parts that have a hierarchical structure, which are loosely related to the graph structure of the wire harness itself. The method estimates the position of each part by using the visual similarity of each part as well as the constraint on the positional relation among the parts imposed by the hierarchical structure. We show the efficacy of this method through several experimental results. wire harness Tohoku University 1 1 c2010 Information

2 % 2 1) 2) 2 2 (1) 1 (2) c2010 Information

3 2.2 (constellation) 1),2) R 1 R 2 3) 4) 3 5) BOF BOF 6) BOF Felzenszwalb 2.3 6) 2 7) 3 3 Felzenszwalb Bag-of-features(BOF) 8) 3 BOF R 1 R 2 R 0 C 2 C c2010 Information

4 G = (V, E) V = {v 1,..., v n } (v i, v j ) E v i BOF v j v i l i l i v 1,..., v n L = {l 1,..., l n } I p(l I) L 1 1 p(l I) p(i L)p(L) (1) L BOF v i I l i I p(i L) = p(i l i ) (2) v i G 7) p(l) G p(l (v i,v j ) E i, l j ) p(l) = (3) v i V p(li)deg(v i) 1 deg(v i) v i G p(l i) v i n p(i L) p(i l i ) p(l i, l j ) (4) n 3. m i (l i ) + d ij (l i, l j ) (5) 3.1 m i (l i ) v i I d ij (l i, l j ) 3 v i i=1 i i=1 (v i,v j ) E (v i,v j ) E 4 c2010 Information

5 v i v j 3.2 (5) v i l i p(i l i ) v i v j l i, l j v i C 8 v i l i m i(l i) R 1 R 2 (v i, v j) E v i v j v i l i v j 4 λ if v j D 1(v i) d ij (l i, l j ) = αλ if v j D 2(v i) (6) v i l i 0 otherwise 3.4 λ m i (l i ) d ij HOG 9) SVM D 1(v i) v i l i LIBLINEAR 10) D 2(v i) D 1(v i) SIFT 11) BOF 8) SVM 0 < α < (5) L 4. l i t i {l (1) i,..., l (t i) i } 4.1 l i t i (5) n t 1 4 i=1 i v i Viterbi 7) v j v i m j(l j) + d ij(l i, l j) v i l i R 0 R 1 R 2 l j l i B j(l i) l i m i(l i) + d ki (l k, l i) + Bj(li) v 4 j k 4.2 l k l i B i(l k ) HOG RGB3 HOG = 3888 R 4 C 2 C5 R R 6 R 0 5 C C 7 7 C 5 C9 0 C 2 R 4 R 1 R 2 C 5 C 5 R 5 C 7 C8 R 6 C 7 0 C 9 5 c2010 Information

5 C 5,, C 8 4 6, R 4, R 5, R 6. 100 400 800 10 3 C 5,, C 8 5 BOF SVM R 1,..., R 6 SIFT SIFT(128 ) (HSV,10 ) 138 BOF 35,..., R 6 6 SVM 1) Fergus, R., Perona, P.

2) Weber, M., Welling, M. and Perona, P.: Unsupervised Learning of Models for false negative false positive Recognition (2000). 3) Agarwal, S. and Roth, D.

, Leonardis, A. and Schiele, B.: Combined Object Categorization and Segmentation With An Implicit Shape Model, In ECCV workshop on statistical 5.

: Example-Based Object Detection in Images by Components, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol.23, pp.349 361 (2001). 6) Fischler, M.A. and Elschlager, R.

6 5 C 5,, C 8 4 6, R 4, R 5, R C 5,, C 8 5 BOF SVM R 1,..., R 6 SIFT SIFT(128 ) (HSV,10 ) 138 BOF 35,..., R 6 6 SVM 1) Fergus, R., Perona, P. and Zisserman, A.: Object class recognition by unsupervised C 8 7,..., R 6 8 HOG SVM , C 8 Viterbi scale-invariant learning, In CVPR, pp (2003). 2) Weber, M., Welling, M. and Perona, P.: Unsupervised Learning of Models for false negative false positive Recognition (2000). 3) Agarwal, S. and Roth, D.: Learning a sparse representation for object detection, 3 Proceedings of the 7th European Conference on Computer Vision, Vol.4, pp (2002). 4) Leibe, B., Leonardis, A. and Schiele, B.: Combined Object Categorization and Segmentation With An Implicit Shape Model, In ECCV workshop on statistical 5. learning in computer vision, pp (2004). 5) Mohan, A., Papageorgiou, C. and Poggio, T.: Example-Based Object Detection in Images by Components, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol.23, pp (2001). 6) Fischler, M.A. and Elschlager, R.A.: The Representation and Matching of Pictorial Structures, Computers, IEEE Transactions on, Vol.100, No.22, pp (1973). 7) Felzenszwalb, P.F. and Huttenlocher, D.P.: Pictorial Structures for Object Recognition, IJCV, Vol.61, p.2005 (2003). 6 c2010 Information

7 8, R 4, R 5, R 6 8) Csurka, G., Dance, C.R., Fan, L., Willamowski, J. and Bray, C.: Visual categorization with bags of keypoints, pp.1 22 (2004). 9) Dalal, N. and Triggs, B.: Histograms of Oriented Gradients for Human Detection, pp (2005). 10) Fan, R.-E., Chang, K.-W., Hsieh, C.-J., Wang, X.-R. and Lin, C.-J.: LIBLINEAR: A Library for Large Linear Classification, Journal of Machine Learning Research, Vol.9, pp (2008). 11) Lowe, D.: Object Recognition from Local Scale-Invariant Features, pp (1999). 7 c2010 Information

(MIRU2008) HOG Histograms of Oriented Gradients (HOG)

(MIRU2008) HOG Histograms of Oriented Gradients (HOG) (MIRU2008) 2008 7 HOG - - E-mail: katsu0920@me.cs.scitec.kobe-u.ac.jp, {takigu,ariki}@kobe-u.ac.jp Histograms of Oriented Gradients (HOG) HOG Shape Contexts HOG 5.5 Histograms of Oriented Gradients D Human