1,a) 1,b) Obstacle Detection from Monocular On-Vehicle Camera in units of Delaunay Triangles Abstract: An algorithm to detect obstacles by using a monocular on-vehicle video camera is developed. Since our main application is when parking a vehicle, we focus on motionless obstacles. Our key idea is to use the knowledge that there are no obstacles along the trajectory of the vehicle so far. The knowledge helps us to find a part of road surface with high probability and to construct a likelihood function. The image is divided into Delaunay triangles so that we can judge each triangle whether it is a part of road surface or not. Keywords: on-vehicle camera, obstacle detection, monocular camera, Delaunay triangulation, Welch s test 1. 1 1-5-1 Gunma University, Kiryu-shi, Gunma, 376 8515 a) nagai@cs.gunma-u.ac.jp b) ohta@cs.gunma-u.ac.jp ( ) 3 (3.5 ) 1
(3.6 ) (4.6 ) 2. [3], [6], [12] [7] [2], [5], [11] [14] [9] [8] [10] 3. 3.1 5 2 2 1 (1) Voodoo 3 : 3 Voodoo[1] 3 ( 3D ) (2) : Voodoo 3D (3) : 3D (Welch s test)[13] (4) : 3D 3D 3D [4] (5) : (3) 1 1 3D (4) 2 2 3.2 Voodoo 3 Voodoo ( ) 3 (3D ) 3D (2 ) 3D Voodoo 3D ( ) Voodoo 2 x z y c = normalize (c[t e ] c[t s ]) (1) z tmp = c v [t] t (2) z = normalize (z tmp (c, z tmp )c) (3) x tmp = d dt c[t s] (4) x = normalize (x tmp (z, x tmp )z) (5) y = z x (6) t s t e 2
1 c[t e] x c[t s] Voodoo c[t] t c v [t] t Voodoo (4) x tmp t s c (, ) normalize() 1 3D v ( ) [x, y, z] t (v c[t s ]) (7) z xy (x,y ) 3.3 Voodoo Voodoo 2 xy 3D z ( ) Voodoo ( 1.4[m]) 3.4 ( 1.4[m]) 3D 1 y 3D z ( ) 1 ( 28[cm] ) 3D z ( μ 1 s 2 1 n 1) 3D z ( μ 2 s 2 2 n 2) 1 2 2 [13] 28[cm] (1) 28[cm] n μ s (2) 1.4[m] 3D (3) 28[cm] 3.5 2 [4] 2 3D 3D 3 3D 3 3D h μ h σh 2 ( z )θ μ θ σθ 2 (h θ 2 ) h θ p(h, θ) h 3
O θ h 2 z 2 h θ h z θ z ( ) h θ p h (h) θ p θ (θ) p(h, a) =p h (h) p θ (θ) (8) p h (h) p θ (θ) *1 { exp( (h μ h) 2 ) (h<μ 2σ p h (h) = h 2 h ) 1 (h μ h ) { exp( (θ μ θ) 2 ) (θ>μ 2σ p θ (θ) = θ 2 θ ) 1 (θ μ θ ) (9) (10) p(h, θ) 3 3D 3D 3D 3.6 3 3D ( ) 3D *1 σ h 0.15 (c) 3 15 30 10 150 75 (c) ( ) 4. 3 3 15 150 4.1 Voodoo 3 3.2 Voodoo 4 4 4 3D 4
7 3D 5 6 4 Voodoo 3D 8 4.2 3.3 4 5 4.3 3.4 5 3D 3D Voodoo 3D 3 6 (8) 5 4.4 3.5 6 3D 5 h θ 5
h θ 3.5 7 2 3D 3 7 45 45 45 45 4.5 3.6 8 3m 6m 9m ( ) 40 22 8 4.6 ( ) 9 9 23 230 100 10 10 110 11 (c) (d) 9 ( 2) 23 30 10 230 78 (c) 155 (d) 110 100 130 31 130 100 6
情報処理学会研究報告 (c) (c) (d) (d) 図 10 図 9 の動画 (その 2) から 100 フレームの短い動画を切り出し て実行 すなわち は 110 番フレームの時点におけるオン ライン実行を想定し 11 から 110 番までの直近 100 フレー ムを用いた結果のうち, 最後のフレームの結果 同様に 図 11 図 10 と同じく 図 9 に示す動画 (その 2) から 100 フレー ム (10 秒) ずつ切り出して実行した結果 図 10 との違いは Voodoo を実際の時系列とは逆向きに実行したことである 道路面の誤判定は殆ど見られない 以降も 20 枚ずつフレームをずらした 100 枚のフレームを用 いた 最後のフレームの結果 道路面の誤判定が見られる 2013 Information Processing Society of Japan 7
(c) 20 100 10 10 (d) Voodoo *2 11 11 10 9 ( 2) 100 10 Voodoo *3 Voodoo 10 11 Voodoo Voodoo Voodoo 5. Voodoo 3 *2 10 100 *3 1 3D Voodoo [1] Voodoo camera tracker. http://www.digilab.unihannover.de/docs/manual.html. [2] Margrit Betke, Esin Haritaoglu, and Larry S. Davis. Real-time multiple vehicle detection and tracking from a moving vehicle, 2000. [3] Christophe Braillon, Cédric Pradalier, James L. Crowley, and Christian Laugier. Real-time moving obstacle detection using optical flow models. In in Proc. of the IEEE Intelligent Vehicle Symp., Tokyo (JP), IEEE, pp. 466 471, 2006. [4] Atsuyuki Okabe, Barry Boots, Kokichi Sugihara, and Sung Nok Chiu. Spatial Tessellations: Concepts and Applications of Voronoi Diagrams. Wiley, 1992. [5] Zehang Sun, Ronald Miller, George Bebis, and David DiMeo. A real-time precrash vehicle detection system. In IEEE International Workshop on Application of Computer Vision, pp. 171 176, 2002. [6] Ashit Talukder and Larry Matthies. Real-time detection of moving objects from moving vehicles using dense stereo and optical flow. IEEE Int. Conf. on Intelligent Robots and Systems (IROS 2004), Vol. 4, pp. 3718 3725, 2004. [7],,.., Vol. 40, No. 2, pp. 441 446, 2009. [8],,.. (PRMU), Vol. 09, pp. 29 36, 2001. [9],,,,,.. (MIRU) 2010 (OS13-3), pp. 1547 1544, 2010. [10],,,,,.. 9 ITS, pp. 121 126, 2010. [11],.., pp. 317 322, 2009. [12],,,,.. 2007-CVIM-158, pp. 41 48, 2007. [13].., 1991. [14],,.., Vol. 75, No. 2, pp. 278 283, 2009. 8