| The playing card auto recognition system is a kind of intelligent sport match and a recreation tool based on image recognition and character recognition technology. In this paper, we gave a comprehensive description on the new algorithms and new technologies relevant to image pre-procession, feature extracting, and character recognition, and the system total design. We also made further research on some key technical problems according to current research state, solving following problems: 1. How to confirm divided recognition region quickly and accurately about card image; 2. How to extract feature value of divided marked image and name character according to characters; 3. How to design multi-layer classifier with high recognition rate.The recognition region division algorithm introduced in this paper uses the mark method of playing cards thoroughly, simplifying the implement of algorithm and fastening the processing speed of image and proposes a new image division method — central coordinate projection method i.e. a two-dimensioned coordinate system whose origin is the centre of image is formed by which the image is projected on level axis and vertical axis. From the result we can confirm the rectangular region including recognition region and designation character,then we divide and recognize the region. By doing so we can not only use machine less,but also improve the confirming speed and accuracy of the playing card, resulting in confirming recognition regions quickly and accurately and implementing the auto division of recognition region in complex background.Moreover, in this paper, we mainly discussed three extraction methods of character vector: image-changing method based on translating, rotation, and unchanging scale; method based on central moment calculation; and method based on character shape and structure.The use of image recognition on playing card auto dealer solves many manual defects in card match successfully. Firstly, the central coordinate projection method may extract recognition region quickly and accurately, then classify marked regions correctly through multi-layer classifie and on this base, we introduce self-designedcard recognition experiment system, which makes base for integral design and practical application in sport match. |
