Chesseye: an integrated framework for accurate and efficient chessboard reconstruction

Abstract

This research paper presents a novel and generalizable approach for precisely detecting and identifying the configuration of pieces on both 2D and 3D chessboard images with different chess sets and varying background contexts. It makes a significant milestone in the digitalization of the chess world by enabling the recreation of physical chess boards on computer screens using a single image. It also provides a framework for real-time tracking and visualization of live chess games using video frames obtained directly from the camera. The novelty lies in the methodology that achieves remarkable accuracy through four key steps: (1) identifying the corner points of the chessboard, (2) detecting the chess pieces, (3) localizing the pieces within the chessboard, and (4) evaluating the position with the best possible variations. The introduction of the Fisher Linear Discriminant Analysis-based dynamic thresholding technique contributes to the perfect 100% accuracy in distinguishing between the white and black chess pieces. The entire algorithm undergoes a thorough experimentation and evaluation process, confirming the effectiveness and versatility of the proposed approach.