Research On Chess Game System Based On Machine Vision

Artifical intelligence(AI)is a very popular research field at present.The theory and methods produced have gradually made outstanding congtributions to all aspects of the economy and society.Therefore,it is of great social significance and value to study the related issues of AI.Machine vision and man-machine game are two important aspects of artificial intelligence.This paper combines them to desin a Chinese chess man-machine game system.The main work is divided into three parts:chess game picture collection,chess location and tecognition algorithm,man-machine game algoithm,and finally realize the interation of the three parts.The first part collects pictures.There are two schemes.First,it is implemented on the embedded development board.The boot progam and the kernel are transplanted,and the file systen is made and installed.On this basis,video driver and application program are used to collect and dispaly chessboard video and capture pictures with screenshot tools.However,the quality of pictures not satisfued.Finally,the mobile phone is used to collet pictures on th built experimental platform.The clarity of the pictures meets the requirements.The second part is chess location recognition algorithm,which can binarize gray image by using gradient modulus as threshold value.Vectors scan from inisde to outside to determine the chessboard area,which has good stability.The proportional method is fast in determing the position coordinates of chess pieces.Aiming at the problem of existence judgment of chess pieces in sub-domain ,Hough transform detection is used to realize it.In order to judge the color of chess pieces,a components in Lab color space is used as a threshold to segment the chess pieces,ande the result is good.The plane rectangular coodinates are teansformed into polar coordinates and projected.Then the projection on the angle axis is processed by Fast Fourier Transform to extract the magnitude feature.The feature obtained by correlation coefficient analysis and screening is better for font classififcation.In order to test the trcognition effect,576 pieces were ollected in the training set and 1024 pieces were collected in the test set.The test results of two sets were as follows:the error rate of chess location and chess color recognition was 0,and the error rate of font recognition was 0.17%and 0.29%.Compared with other non-screening methods,the error rate of font recognition was reduced by 0.35%and 0.39%respectively.The third part is the man-machine game algorithm,in which search technology and evaluation function are two key aspects to improve the level of computer game.After analysis,tailoring and transplantation,the code is transplanted into the new Visual C++ MFC.Finally,by making library files and so on,the three parts of picture collection,chess location recognition and human-computer game are integrated into one system.Man-machine game verification of the whole chess game is carried out,and the operation is stable and normal.This paper deals with the research results of machine vision and human-machine game,and can also be transferred to other scenarios,such as parts defect recognition,medical image recognition,etc.It has certain theoretical and practical significance.