仿人足球机器人快速多目标识别方法  

Fast Multi-Target Recognition Method for Humanoid Robot Playing Soccer

作　　者：吴峰华[1] 杨哲海 张玉萍 王昊 刘实 尹竞瑶 金鑫 王承业 盖宇 Wu Fenghua;Yang Zhehai;Zhang Yuping;Wang Hao;Liu Shi;Yin Jingyao;Jin Xin;Wang Chengye;Gai Yu(Department of Robot Engineering,Shenyang City University,Shenyang 110112)

机构地区：[1]沈阳城市学院机器人工程系

出　　处：《计算机辅助设计与图形学学报》2019年第12期2152-2165,共14页Journal of Computer-Aided Design & Computer Graphics

基　　金：辽宁省自然科学基金(1557901922092);沈阳市科技创新平台建设计划市级重点实验室项目(18-007-0-10);沈阳市高层次创新人才项目(RC180375)

摘　　要：针对全自主仿人足球机器人在自身动态特性显著和比赛场地复杂环境条件下快速准确识别多种目标物的要求,采用融合形态学分析的类Haar特征构造热图的方法对机器人视域内的疑似目标物进行粗略选取,形成“感兴趣区域”.再构造轻量级Tiny-DNN卷积神经网络对“感兴趣区域”进行快速分类,准确判断这些区域是否包含目标物.机器人运行过程中自动生成的大量“感兴趣区域”作为卷积神经网络的离线训练过程正负样本集,避免了在海量图片上人为标注目标物样本的大量工作与耗时及不确定性偏差.该方法应用于搭载CPU处理器的SYCU-Legendary仿人机器人,使其能够在0.03 s内对于目标物足球、球门柱和罚球点的识别率分别达到95.8%,96.2%和96.0%.目标识别过程受比赛场地异物和光线变化影响较小,夺得了2018年和2019年足球机器人世界比赛(RoboCup)中国赛区冠军.实际应用结果表明该方法具有较高的推广应用价值.Fully autonomous humanoid robot soccer competition requires the dynamic-featured robot recognizing multi-target both accurately and rapidly under complex field environment.A fast multi-target recognition method for humanoid soccer robot is proposed.Haar-like features integrated with morphological analysis is employed to construct heat maps,on which the candidate patches supposed containing object in the robot’s field of view are roughly selected to form a set of“region of interest(ROI)”.This ROI set is quickly classified into target or non-target regions through a lightweight Tiny-DNN convolution neural network.Meanwhile,each ROI generated automatically during the robot operation can be directly collected and stacked into positive or negative sample sets for the off-line training of the convolution neural network,thus avoiding tremendous labor work and uncertainty deviation of manually tagger of target sample on quantities of pictures.The proposed method is applied to CPU supported SYCU-Legendary humanoid robot,enabling it to recognize 95.8%,96.2%and 96.0%of football,goalposts and penalty marks within 0.03 seconds.Besides,foreign matters and light changes cast little influence on targets recognition so that SYCU-Legendary team won the championship in 2018 and 2019 RoboCup China Open,which indicates the method in this paper is worthy to be widely extended to other application domains.

关 键 词：仿人机器人 形态学类Haar特征 热图 Tiny-DNN卷积神经网络 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]