YGSM.zip_fusion_双目导航_人工智能/神经网络/深度学习

<html xmlns="http://www.w3.org/1999/xhtml"> <head> <meta charset="utf-8"> <meta name="generator" content="pdf2htmlEX"> <meta http-equiv="X-UA-Compatible" content="IE=edge,chrome=1"> <link rel="stylesheet" href="https://static.pudn.com/base/css/base.min.css"> <link rel="stylesheet" href="https://static.pudn.com/base/css/fancy.min.css"> <link rel="stylesheet" href="https://static.pudn.com/prod/directory_preview_static/6252cffd6caf5961927a6e15/raw.css"> <script src="https://static.pudn.com/base/js/compatibility.min.js"></script> <script src="https://static.pudn.com/base/js/pdf2htmlEX.min.js"></script> <script> try{ pdf2htmlEX.defaultViewer = new pdf2htmlEX.Viewer({}); }catch(e){} </script> <title></title> </head> <body> <div id="sidebar" style="display: none"> <div id="outline"> </div> </div> <div id="pf1" class="pf w0 h0" data-page-no="1"><div class="pc pc1 w0 h0"><img class="bi x0 y0 w1 h1" alt="" src="https://static.pudn.com/prod/directory_preview_static/6252cffd6caf5961927a6e15/bg1.jpg"><div class="t m0 x1 h2 y1 ff1 fs0 fc0 sc0 ls0 ws0">第 35 卷 第 9 期</div><div class="t m0 x2 h2 y2 ff2 fs0 fc0 sc0 ls0 ws0">2 0 0 3 年 9 月</div><div class="t m0 x3 h3 y3 ff3 fs0 fc0 sc0 ls0 ws0"> </div><div class="t m0 x4 h4 y4 ff1 fs1 fc0 sc0 ls0 ws0">哈 尔 滨 工 业 大 学 学 报</div><div class="t m0 x5 h5 y5 ff2 fs1 fc0 sc0 ls0 ws0">JOURNAL OF HARBIN INSTITUTE OF TECHNOLOGY</div><div class="t m0 x6 h6 y6 ff3 fs1 fc0 sc0 ls0 ws0"> </div><div class="t m0 x7 h7 y7 ff2 fs0 fc0 sc0 ls0 ws0">Vol1 35 No19</div><div class="t m0 x8 h7 y8 ff2 fs0 fc0 sc0 ls0 ws0">Sep. , 2003</div><div class="t m0 x9 h3 y9 ff3 fs0 fc0 sc0 ls0 ws0"> </div><div class="t m0 xa h8 ya ff5 fs2 fc0 sc0 ls0 ws0">基于异构双目视觉的全自主足球机器人导航</div><div class="t m0 xb h9 yb ff6 fs3 fc0 sc0 ls0 ws0">高庆吉, 洪炳熔, 阮玉峰</div><div class="t m0 xc ha yc ff2 fs4 fc0 sc0 ls0 ws0">( 哈尔滨工业大学 计算机科学与技术学院, 黑龙江 哈尔滨 150001)</div><div class="t m0 xd ha yd ff5 fs4 fc0 sc0 ls0 ws0">摘 要: 通过理论分析和实验, 论述了全自主足球机器人异构双目视觉协调导航的合理 性和有效 性, 给出 了</div><div class="t m0 xd hb ye ff7 fs4 fc0 sc0 ls0 ws0">测量目标相对坐标的模型, 分析了基于双目协调的足球 机器人导航数据融合方法, 提出了双目协调导航模 型</div><div class="t m0 xd hb yf ff7 fs4 fc0 sc0 ls0 ws0">及基于行为的目标实现策略. 在所开发的小型全自主足 球机器人上的仿真和实验结果, 证明了双目协调的 有</div><div class="t m0 xd hb y10 ff7 fs4 fc0 sc0 ls0 ws0">效性和实用性.</div><div class="t m0 xd ha y11 ff5 fs4 fc0 sc0 ls0 ws0">关键词: 异构双目; 全自主足球机器人; 导航; 数据融合</div><div class="t m0 xd ha y12 ff5 fs4 fc0 sc0 ls0 ws0">中图分类号: TP24 文献标识码: A 文章编号: 0367- 6234( 2003) 09- 1029- 04</div><div class="t m0 xe hc y13 ff8 fs3 fc0 sc0 ls0 ws0">Navigation of smal-l size autonomous soccer robot based on</div><div class="t m0 xf hc y14 ff8 fs3 fc0 sc0 ls0 ws0">iso- vision system with two cameras</div><div class="t m0 x10 h5 y15 ff2 fs1 fc0 sc0 ls0 ws0">GAO Qing- ji, HONG Bing- rong, RUAN Yu- feng</div><div class="t m0 x11 hd y16 ff2 fs4 fc0 sc0 ls0 ws0">( School of Computer Science and Technology, Harbin Institute of Technology , Harbin 150001, China)</div><div class="t m0 x12 h5 y17 ff8 fs1 fc0 sc0 ls0 ws0">Abstract: The rationality and effectiveness of coordinated navigation of iso- vision system with two cameras for soccer</div><div class="t m0 x13 h5 y18 ff2 fs1 fc0 sc0 ls0 ws0">robot is discussed through theoretical analysis and experiments. Models to get the local coordinates are given. The</div><div class="t m0 x12 h5 y19 ff2 fs1 fc0 sc0 ls0 ws0">approach of data fusion and coordination based on the iso- vision with two cameras to implement navigat ion of soccer</div><div class="t m0 x13 h5 y1a ff2 fs1 fc0 sc0 ls0 ws0">robot are analyzed. The validation and practicability of navigation based on coordination of iso-vision with two cam-</div><div class="t m0 x13 h5 y1b ff2 fs1 fc0 sc0 ls0 ws0">eras are proved through experiments on the accomplished robot.</div><div class="t m0 x12 h5 y1c ff8 fs1 fc0 sc0 ls0 ws0">Key words: iso-vision with two cameras; autonomous soccer robot ; navigation; data fusion</div><div class="t m0 x13 h2 y1d ff5 fs0 fc0 sc0 ls0 ws0">收稿日期: 2003- 06- 20.</div><div class="t m0 x13 h2 y1e ff5 fs0 fc0 sc0 ls0 ws0">基金项目: 国家高技术研究发展计划资助项目( 863- 2001AA422270) .</div><div class="t m0 x13 h2 y1f ff5 fs0 fc0 sc0 ls0 ws0">作者简介: 高庆吉( 1966- ) , 男, 博士研究生, 副教授;</div><div class="t m0 x14 h2 y20 ff1 fs0 fc0 sc0 ls0 ws0">洪炳熔( 1937- ) , 男, 博士, 教授, 博士生导师.</div><div class="t m0 x15 h4 y21 ff3 fs1 fc0 sc0 ls0 ws0"> 目前, 足球机器人视觉系统有单目固定视觉、</div><div class="t m0 x12 h4 y22 ff1 fs1 fc0 sc0 ls0 ws0">单目活动视觉</div><div class="t m0 x16 he y23 ff2 fs5 fc0 sc0 ls0 ws0">[ 1]</div><div class="t m0 x17 h4 y22 ff1 fs1 fc0 sc0 ls0 ws0">和双目活动视觉结构. 仿人眼的</div><div class="t m0 x12 h4 y24 ff1 fs1 fc0 sc0 ls0 ws0">对称式结构的双目视觉系统通过图像匹配可以获</div><div class="t m0 x12 h4 y25 ff1 fs1 fc0 sc0 ls0 ws0">得目标的深度信息</div><div class="t m0 x18 he y26 ff2 fs5 fc0 sc0 ls0 ws0">[ 2, 3]</div><div class="t m0 x19 h4 y25 ff2 fs1 fc0 sc0 ls0 ws0">, 但这种结构 的视野范围</div><div class="t m0 x12 h4 y27 ff1 fs1 fc0 sc0 ls0 ws0">有局限性. 采用非对称结构的视觉系统至少有 1</div><div class="t m0 x12 h4 y28 ff1 fs1 fc0 sc0 ls0 ws0">个摄像机具有 1 个以上的自由度, 可通过检测全</div><div class="t m0 x12 h4 y29 ff1 fs1 fc0 sc0 ls0 ws0">角度反光凸镜中的图像获得全方位视野, 再通过</div><div class="t m0 x12 h4 y2a ff1 fs1 fc0 sc0 ls0 ws0">另外 1 个摄像机对重点目标进行注视跟踪</div><div class="t m0 x1a he y2b ff2 fs5 fc0 sc0 ls0 ws0">[ 4]</div><div class="t m0 x1b h4 y2a ff2 fs1 fc0 sc0 ls0 ws0">, 但</div><div class="t m0 x12 h4 y2c ff1 fs1 fc0 sc0 ls0 ws0">这种结构要占用机器人的高度空间, 不适用于有</div><div class="t m0 x12 h4 y2d ff1 fs1 fc0 sc0 ls0 ws0">高度约束情况.</div><div class="t m0 x1c h4 y2e ff1 fs1 fc0 sc0 ls0 ws0">本文针对 FIRA 的 RoboSot 比赛规则, 设计开</div><div class="t m0 x12 h4 y2f ff1 fs1 fc0 sc0 ls0 ws0">发了参赛机器人, 采用异构双目活动视觉系统, 即</div><div class="t m0 x12 h4 y30 ff1 fs1 fc0 sc0 ls0 ws0">一个固定摄像机和一个可以水平旋转的摄像机,</div><div class="t m0 x12 h4 y31 ff1 fs1 fc0 sc0 ls0 ws0">通过合理的资源分配及协调机制使系统在视野范</div><div class="t m0 x1d h4 y21 ff1 fs1 fc0 sc0 ls0 ws0">围、测距精度及处理速度方面达到最佳匹配.</div><div class="t m0 x1e hf y32 ff2 fs3 fc0 sc0 ls0 ws0">1 异构双目视觉系统</div><div class="t m0 x1f h4 y33 ff1 fs1 fc0 sc0 ls0 ws0">异构双目视觉系统建立在一种全自主的机器</div><div class="t m0 x1d h4 y34 ff1 fs1 fc0 sc0 ls0 ws0">人体系结构上. 异构双目摄像机分别安装在机器</div><div class="t m0 x1d h4 y35 ff1 fs1 fc0 sc0 ls0 ws0">人的顶部和中下部, 如图 1 所示.</div><div class="t m0 x20 ha y36 ff1 fs4 fc0 sc0 ls0 ws0">图 1 异构双目视觉布局图</div><div class="t m0 x21 h10 y37 ff2 fs4 fc0 sc0 ls0 ws0">Fig. 1 Location of iso- vision system</div><div class="t m0 x22 h4 y38 ff3 fs1 fc0 sc0 ls0 ws0"> 基于颜色和形状相结合的方法可以确定出单</div><div class="t m0 x1d h4 y39 ff1 fs1 fc0 sc0 ls0 ws0">幅图像中是否存在目标以及目标的屏幕坐标. 对</div><div class="t m0 x1d h4 y3a ff1 fs1 fc0 sc0 ls0 ws0">于平坦的地面环境, 摄像机的光心与地面的垂直</div><div class="t m0 x1d h4 y3b ff1 fs1 fc0 sc0 ls0 ws0">距离可测, 其光轴与地平面的夹角可间接测得, 因</div></div><div class="pi" data-data='{"ctm":[1.677504,0.000000,0.000000,1.677504,0.000000,0.000000]}'></div></div> </body> </html>