Chess game recognition method and device, equipment and storage medium

Abstract

The embodiment of the invention discloses a chess game recognition method and device, equipment and a storage medium. The method comprises the steps of determining a first group of pixel coordinates of each corner point on a to-be-processed chessboard image; based on the first group of pixel coordinates of each corner point, determining a second group of pixel coordinates of each corner point on each corresponding point on a chess piece plane, wherein the chess piece plane is determined based on the surface of at least one chess piece on the chessboard image; performing perspective transformation on the chessboard image based on the first group of pixel coordinates of each angular point and the second group of pixel coordinates of each corresponding point to obtain two corresponding view angle correction images, wherein the two view angle correction graphs are respectively used for representing the actual positions of the upper and lower surfaces of the chess pieces in the chessboard image; and determining an identification result of the chessboard image based on the two view angle correction graphs.

Description

technical field [0001] This application relates to the field of computer vision, and involves but is not limited to chess game recognition methods, devices, equipment, and storage media. Background technique [0002] Vision-based chess game recognition is an important part of chess robot software, and its core work is to accurately classify and coordinate the chess pieces on the chessboard. [0003] However, the existing chess piece identification and positioning methods are all for processing the pictures taken by the camera perpendicular to the plane of the chessboard, which greatly limits the selection of the installation position of the camera, so that the mechanical structure of the whole chess-playing robot cannot be flexibly designed. On the one hand, most of the existing methods use traditional image processing methods, which have high complexity, poor generalization and low accuracy, which cannot meet the actual needs; on the other hand, target detection algorithms ...

AI Technical Summary

Problems solved by technology

[0003] However, the existing chess piece identification and positioning methods are all for processing the pictures taken by the camera perpendicular to the plane of the chessboard, which greatly limits the selection of the installation position of the camera, so that the mechanical structure of the whole chess-playing robot cannot be flexibly designed.

On the one hand, most of the existing methods use traditional image processing methods, which are high in complexity, poor in generalization and low in accuracy, and cannot meet actual needs; on the other hand, target detection algorithms based on deep learning can only obtain image The detection frame is a rectangular frame, and the imaging of the chess pieces on the image plane with an oblique angle is not a standard circle, so it is impossible to directly use the results of the target detection algorithm to accurately locate the center of the chess pieces

Images