A Machine Vision System for Chinese Chess-Playing Robot

Abstract

This paper introduces a machine vision system for Chinese chess-playing robots which is usually regarded as a form of recreation. The machine vision system with two color cameras takes simultaneously two images of a chessboard and round pieces on the chessboard from different angles (views). Firstly, original images are handled by a series of image processing operations such as color conversion, binarization and denoise. Secondly, a hierarchical Hough transform algorithm was taken to detect lines and circles in the binarized image. Circles with reasonable radii nearly centered on the crossings of chessboard will be considered as pieces on the chessboard and its corresponding Chinese character inside the circle would be recognized based on BP neural network and ring intersection points. The 3D coordinates of pieces can be calculated through computation and then transformed to coordinates expressed by row and column. Finally, results of image processing and pattern recognition procedures will be sent to robot control system to manipulate end-effector of a robot to move round pieces from one place to another desired. Experimental results reveal that the designed machine vision system in this paper can work well with higher reliability.