社団法人人工知能学会 Japanese Society for Artificial Intelligence 人工知能学会研究会資料 JSAI Technical Report SIG-Challenge-B3 (5/5) RoboCup SSL Humanoid A Proposal and its Application of Color Voxel Server for RoboCup SSL Humanoid Yasuo UEDA Yasuhiro MASTANI Osaka Electro-Communication University Abstract In this paper, the authors propose color voxel server as a shared vision system for RoboCup SSL Humanoid, which generates coloring voxel data of 3D space on a field from images taken with multiple external cameras and supplies it via network. Kimura proposed an original voxel server in 21, which is insufficient to distinguish objects. Because it generates binary voxel data. Therefore the proposed server gives color information to voxel without spoiling realtime performance by not using full color and reducing number of colors in the camera images. Seitz s of voxel coloring is used to solve color of voxel. It takes about 8[ms] for a prototype program to generate 4-color 1[mm]-cubic voxels of 45[mm] 325[mm] 5[mm] space from 8 camera images and supply compressed data via network. Moreover the authors show that it is possible to distinguish objects and decide action of based on color voxel data provided from the server. 1 RoboCup (Small Size Robot League, SSL)[, 211] SSL Humanoid[, 21] Figure 1 Figure 1: Initial phase of the SSL Humanoid (29) Figure 2: Final phase of the SSL Humanoid (215) 2 SSL Humanoid Figure 2 215 3 SSL 21 SSL-Vision SSL-Vision 9
1 Table 1: Identification by color of voxel Voxel Mode of expression Nothing Other 1 Orange 2 Blue 3 Yellow 4 SSL Humanoid SSL-Vision 3 3 [, 21] 8 325[mm] 45[mm] 5[mm] 1[mm] [, 211] 2 2 bit Table 1 5 2 2.1 2.1.1 HSV 2.1.2 Space Carving Method(SCM) SCM ( ) 2.1.3 Voxel Coloring [M.Seiz, 1997] 3 z 1 1 2.2 UDP 1 [mm] 3 [mm] [byte] 1 2byte 3[bit] () 13[bit] 1
11 Figure 5: Silhouette images Figure 3: Camera layout Table 2: Specifications of the experimental PCs Server PC OS Windows 7 Professional 64bit Main Mamory 5GByte CPU Intel Core i7 95(3.19GHz) Graphic Board NVIDIA GeForce GTX26 Client PC OS Windows 7 Professional 64bit Main Mamory 4GByte CPU Intel Core i7 X94(2.13GHz) Graphic Board ATI Mobility Radeon HD 58 Series Figure 6: Generated voxel data (a) Real image (b) Voxel atea Figure 4: Camera images 1 Figure 7: Camera image and voxel data proposed Kimura s 2.3 PC Table 2 Figure 3 Voxel Coloring Figure 3 4 z 3 Figure 4 Figure 5 Figure 67 Figure 5 Figure 8 4 time[ms] 8 6 4 2 1 2 3 4 5 6 7 8 9 1 num Figure 8: Comparison of the processing time of the proposed and Kimura s 34% 3 11
12 Figure 9: Two types of labeling Figure 1: Initial positions of and 3.1 1 2 Figure 9 3.2 Voxel Coloring 3.3 3.4 3.5 18[deg] 15 1 5-5 -1-15 -15-1 -5 5 1 15 Figure 11: Trajectories of and by Kimura s 15 1 5-5 -1-15 -15-1 -5 5 1 15 Figure 12: Trajectories of and by the proposed 4 PC Table 2 4.1 2 12
13 15 1 5-5 -1-15 35 4 45 5 55 6 Figure 16: Initial and target positions of s Figure 13: X coordinate of and by Kimura s 15 1 2 15 1 1 5 5-5 -5-1 -15-1 35 4 45 5 55 6 Figure 14: X coordinate of and by the proposed -15-2 -15-1 -5 5 Figure 17: Trajectories of s time[ms] 16 14 12 1 8 6 4 2 proposed Kimura s 1 1[ms] 2 5 1 15 2 25 3 35 num Figure 15: Comparison of the processing time of the proposed and Kimura s by client Figure 1 Figure 11 x Figure 13 Figure 12 x Figure 14 Figure 15 53[s] 4.2 Figure 16 2 2 1 Figure 17 x y Figure 18 19 Figure 2 ( ) Figure 19 24[s] 1 2 Figure 2 1 1 13
14 5-5 -1 1 2 1 flag 2 flag decision flag -15 action decision -2 2 25 3 35 4 45 5 55 Figure 18: X coordinate vs. time of 1, 2, and 15 1 5-5 -1 1 2 1 flag 2 flag decision flag action -15 decision -2 2 25 3 35 4 45 5 55 Figure 19: Y coordinate vs. time of 1, 2, and theta[deg] 15 1 5-5 -1-15 -2-25 1 2 1 flag 2 flag decision flag 2 25 3 35 4 45 5 55 action decision Figure 2: Direction vs. time of 1, 2, and 1 36[s] 1 2 2 1 38[s] 1 1 43[s] flags flags flags 1 2 5 RoboCup SSL Humanoid [, 211] : 3, vol.52 No.1, pp.95 11, 211. [, 21] : RoboCup Soccer SSL Humanoid,, Vol.25, No.2, pp.213 219, 21. [, 21] : 3 RoboCupSoccer SSL Humanoid 3, 31 AI, pp.33 38, 21. [, 211] :, 33 AI, pp.7 12, 211. [M.Seiz, 1997] Steven M.SeizCharles R.Dyer: Photorealistic Scene Reconstruction by Voxel Coloring Computer Vision and Pattern Recognition Conf pp.167 1731997 14
41 Figure 4: 2 Figure 6: 2 41