基于卷积神经网络的机器人象棋识别与视觉定位算法设计  

Chess Recognition and Visual Localization Algorithm Based On Convolution Neural Network

作　　者：熊勋 尹玲 陈新度 张斐 吴鹏[1] 王磊杰[1] XIONG Xun;YIN Ling;CHEN Xindu;ZHANG Fei;WU Peng;WANG Leijie(School of Electrical Engineering and Intelligentization,Dongguan University of Technology,Dongguan 523808,China)

机构地区：[1]东莞理工学院机械工程学院

出　　处：《东莞理工学院学报》2019年第5期104-109,共6页Journal of Dongguan University of Technology

基　　金：广东省省级科技计划项目(2016B090912002);广东省智能制造技术与装备企业重点实验室项目(2017B030302004);广东省自然科学基金项目(2016A030313133);广东省普通高校机器人与智能装备重点实验室(2017KSYS009;广东省普通高校特色创新项目:面向大数据应用的车间装备智能化关键技术及应用开发(2017KTSCX176);佛山高新区专用装配创新型产业集群建设(2015B010102005);广东省制造装备数字化重点实验室开放运行专项项目(2017B030314146)

摘　　要：针对机器人下棋时要求识别准确、和快速响应的要求,提出了一种基于卷积神经网络的象棋识别与视觉定位算法。通过对目标图像一系列预处理算法,经由Hough圆提取算法优化,就能准确的定位到棋子区域,运用已经训练好的卷积神经网络模型对目标进行识别,为了验证识别算法的优越性,将棋子进行各个方向的旋转,实验结果表明所使用的定位算法精度达到0.17 mm,识别精度保持在98%以上,识别整副棋子用时3.63 s,算法识别性能优于传统机器视觉。Aiming at the requirement of the robot to recognize the accuracy and the fast response of the robot, a Chinese chess recognition and vision positioning algorithm based on the convolution neural network is proposed.Through a series of pre-processing algorithms on the target image, the algorithm can be accurately positioned in the area of the pawn through the Hough circle extraction algorithm, and the target is identified by using the trained convolution neural network model, and the chess pieces are rotated in various directions in order to verify the advantages of the recognition algorithm. The experimental results show that the accuracy of the positioning algorithm is 0.17 mm, the recognition accuracy is above 98%, the identification of the whole piece is 3.63 seconds, and the identification of the algorithm can be superior to the traditional machine vision.

关 键 词：象棋识别 视觉定位 卷积神经网络 机器视觉 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]