Research And Implementation Of Chess Robot Based On Global Vision

At present,most of the human-computer games stay in the human-computer interaction,if the user-robot game can be realized,the experience and entertainment of machine intelligence can be more manifested.In this paper,a global vision-based chess robot system was constructed,and chessboard inpainting,detection and recognition,and chess piece localization algorithms were focused on.The main work was as follows:An improved chessboard inpainting algorithm based on DC-GAN was proposed to solve the problem of partial missing of chessboard images caused by shadows and reflections.Experimental results showed that the U-Net network was used to replace the traditional generator structure to construct reconstruction loss and adversarial loss functions.On the three self-made artificial noise data sets,the network performed well,and the regional context was restored to achieve consistency and coherence.A chessboard segmentation algorithm with adaptive thresholding combined with morphological operations,a chessboard vertex detection algorithm based on block convex hull,and a chessboard correction algorithm based on perspective transformation were proposed to solve the problem of chessboard automatic detection caused by complex background and distortion.A chessboard type recognition algorithm with custom board pattern matching was proposed to solve the problem of undetermined board types.The experimental results showed that the board can automatically complete the detection and identification type operation.Focused on the problem of chess piece positioning error under uneven illumination distribution,a general chess piece positioning algorithm based on image mask was proposed.The experimental results demonstrate that,the positioning accuracy of Go and Chinese chess pieces arrives by 95.5%and 99.06%respectively under four kinds of uneven illumination.And the inaccurate local positioning problems of chess pieces caused by chess pieces adhesion,chess pieces projection and lens distortion are solved.Combining the above algorithms,a highly flexible and versatile chess robot system supporting multiple chess types was designed.The communication and functional design of the vision system,the drop subsystem,the game system and the real-time display system were implemented,and functions such as saving games and simulcasting games were also included.