Era of Virtual Picture from Various Perspective Point Picture Database.


52 views
Uploaded on:
Category: General / Misc
Description
・. Era of Virtual Picture from Various Perspective Point Picture Database . Haruki Kawanaka, Nobuaki Sado and Yuji Iwahori Nagoya Foundation of Innovation, Japan . Foundation.
Transcripts
Slide 1

・ Generation of Virtual Image from Multiple View Point Image Database Haruki Kawanaka, Nobuaki Sado and Yuji Iwahori Nagoya Institute of Technology, Japan

Slide 2

Background The soccer playing diversion has gotten to be famous in Japan following the World Cup Soccer of Japan and South Korea cosponsorship was held in 2002. It is wanted to see the amusement from different perspective focuses. setting cameras at the opposite side of the objective at the roof to view down. numerous cameras are set at different areas a camera has the capacity of pan−tilt−zoom ・ But, these are not free perspectives.

Slide 3

Previous strategy to create a virtual picture ・ expansive scale environment with numerous camera settings and establishments It takes much cost. the application is confined at just that stadium. utilizing a couple of cameras and a movement catch at just the indoor space the exceptional wear and a few markers ・ It is hard to use in genuine diversions.

Slide 4

Present methodology ・ The back\'s names number are created as the virtual picture. The stance of every player is not considered.

Slide 5

Study Purpose ・ produce a virtual picture at another perspective point from a genuine picture without the extraordinary environment

Slide 6

Proposed Method The suitable stance picture of every player is resolved from utilizing different perspective picture database of a player’s CG model. Every stance picture is integrated at the position to the virtual scene. The position of every player is thought to be given by the direction framework.

Slide 7

Trajectory of two players Trajectory Recording System We have created Trajectory Recording System. ・

Slide 8

Flow of Proposed Method Creation of database by CG model Generation of virtual picture for every player from picture database Recognize the posture of a player Generate the comparing virtual picture from another perspective point Synthesis of another perspective point picture ・

Slide 9

Creation of database by CG model ~ Images for Database ~ Image Database (Human model) is made by CG. Different movements ( run, walk, shot, pass, heading, trap and so on ) 280 postures From 8 view bearing (revolution with each 45â°) absolute 2240 pictures ・

Slide 10

Creation of database by CG model ~ Processing of Each Image ~ CG Image of human model To dispose of numerous components, for example, the state of light source, skin shading, hair and uniform… It is important to spare the information size and the inquiry time of the picture database. Picture database is made utilizing the outline. This relies on upon the distinction of stance however does not rely on upon such element of every player. Outline ・

Slide 11

N Creation of database by CG model ~ Processing of Each Image ~ Normalize the picture size 1-dimensional information The picture size for every stance is standardize. The rectangle district which encompasses the outline of every stance, is separated. The removed district just touches to the square with keeping its angle proportion. By the raster examine, one dimensional development of the standardized picture is made. [00011000 ・・・ ] N×N ・

Slide 12

[00011000 ・・・ ] [01011010 ・・・ ] ・・・ [10001001 ・・・ ] Creation of database by CG model ~ Principal Component Analysis ~ ・ Set of picture information (all stances & all perspective) Image information If the entirety of eigen qualities gets to be more than 90% powerful ・ Compress & Projection onto eigen space ・ Covariance network ・ eigen values, eigen vectors ・

Slide 13

Generation of virtual picture for every player ~ Recognize the posture of a player(1) ~ Detection of player N by eigen vector 1-D [0000111 ・・・ ] N Projection onto eigen space Normarize ・ Each outline is standardized, changed to one dimensional vector and anticipated to a point in the eigen space. ・

Slide 14

Generation of virtual picture for every player ~ Recognize the posture of a player(2) ~ least separation A B e 3 given picture e 2 e 1 Result of hunt CG picture Real picture Eigen Space ・ When An is given, B is chosen as the most comparative example. The stance of picture An is perceived as B . ・

Slide 15

Generation of virtual picture ~ Generate the comparing virtual picture from another perspective point ~ Another perspective picture is produced using aftereffect of hunt as indicated by perspective point & view bearing of virtual picture. ・ + Coordinates are gained from direction recording framework ( x , y ) ・ Generated virtual scene + ・ Virtual stadium made utilizing the OpenGL ・

Slide 16

Experiments Actual unique picture Virtual picture from the same perspective heading as unique ・ The examination of stance acknowledgment ・

Slide 17

Experiments Virtual picture from distinctive perspective point Original picture Player’s position is altered. Perspective is moved. ・ Texture is utilized as soccer field. ・ It is additionally conceivable to produce a liveliness of motion picture by interfacing every edge picture consecutively.

Slide 18

Conclusion another way to deal with produce a virtual picture from another perspective point is proposed. Multi-picture database to apply the eigen space technique for the posture acknowledgment. This methodology is basic however creates the sensible virtual scene. ・

Slide 19

Future Works It is hard to separate the total posture of every player. It is additionally troub

Recommended
View more...