Bridge Card Recognition Based On Deep Neural Network

Bridge game is a competitive intellectual game. It is also an official event in many athletic games. The bridge game can be broadcast live on Internet and TV by recognizing and recording each bridge card played during the game. Based on computer vision theories, image recognition can be utilized to recognize bridge cards and substitute human recognition, resulting in an automatic and digital broadcast of the game. Deep neural networks gain its popularity due to its good representative ability and successes on many image recognition tasks. In this paper, we investigate the principle of deep neural networks and apply it to bridge card recognition. Our recognition scheme consists of two parts:card detection and card recognition. First of all, we design some special symbols which are very easy to be detected in the image. We paste these symbols right beside the bridge cards during the game. As long as we locate these symbols we locate the cards. Then, a convolutional neural network is utilized to classify the cards. Our method achieved99.896%recognition rate which is very close to human recognition rate. Hence, our system can be applied to practical use.