• lgz
    了解作者
  • matlab
    开发工具
  • 874KB
    文件大小
  • zip
    文件格式
  • 0
    收藏次数
  • 1 积分
    下载积分
  • 3
    下载次数
  • 2018-12-07 10:35
    上传日期
一种基于环投影矢量变换的多尺度旋转图像模板匹配方法
A Template Matching Method.zip
  • A Template Matching Method for Multi-Scale and Rotated Images Using Ring Projection Vector Conversion.pdf
    898.9KB
内容介绍
<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/625a50dabe9ad24cfae94f30/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/625a50dabe9ad24cfae94f30/bg1.jpg"><div class="t m0 x1 h2 y1 ff1 fs0 fc0 sc0 ls0 ws0">A Template Match<span class="_ _0"></span>ing Method for Mult<span class="_ _0"></span><span class="ls1 ws1">i-Scale and R<span class="_ _0"></span>otated <span class="_ _0"></span>Image<span class="_ _0"></span>s Using R<span class="_ _0"></span>ing </span></div><div class="t m0 x2 h2 y2 ff1 fs0 fc0 sc0 ls2 ws2">Projection Vector Conversion<span class="_ _0"></span> </div><div class="t m0 x3 h3 y3 ff2 fs1 fc0 sc0 ls0 ws0"> </div><div class="t m0 x4 h4 y4 ff2 fs2 fc0 sc0 ls3 ws3">Xinwei Q<span class="_ _0"></span>i </div><div class="t m0 x5 h3 y5 ff2 fs1 fc0 sc0 ls4 ws4">College of Information Science and Engineering </div><div class="t m0 x6 h3 y6 ff2 fs1 fc0 sc0 ls5 ws5">Northeastern University </div><div class="t m0 x7 h3 y7 ff2 fs1 fc0 sc0 ls5 ws5">Shenyang, China </div><div class="t m0 x8 h3 y8 ff2 fs1 fc0 sc0 ls6 ws6">e-mail: xinwei7@outlook<span class="_ _0"></span>.com<span class="_ _1"></span> </div><div class="t m0 x9 h4 y4 ff2 fs2 fc0 sc0 ls7 ws7">Ligang Miao </div><div class="t m0 xa h3 y5 ff2 fs1 fc0 sc0 ls4 ws4">College of Information Science and Engineering </div><div class="t m0 xb h3 y6 ff2 fs1 fc0 sc0 ls5 ws5">Northeastern University </div><div class="t m0 xc h3 y7 ff2 fs1 fc0 sc0 ls8 ws8">Qinhu<span class="_ _0"></span>angdao, China </div><div class="t m0 xd h3 y8 ff2 fs1 fc0 sc0 ls4 ws9">e-mail: miaolg@hotm<span class="_ _0"></span>ail.co<span class="_ _1"></span>m</div><div class="t m0 x3 h3 y9 ff2 fs1 fc0 sc0 ls0 ws0"> </div><div class="t m0 x3 h3 ya ff2 fs1 fc0 sc0 ls0 ws0"> </div><div class="t m0 xe h5 yb ff3 fs3 fc0 sc0 ls9 ws0">Abstract<span class="ff4 ls0">&#8212;<span class="ff1 lsa wsa">Template matching is a pro<span class="_ _0"></span>cess of finding template </span></span></div><div class="t m0 xe h5 yc ff1 fs3 fc0 sc0 lsb wsb">image lo<span class="_ _1"></span>catio<span class="_ _1"></span>n fr<span class="_ _1"></span>om a <span class="_ _1"></span>know<span class="_ _1"></span>n imag<span class="_ _1"></span>e, w<span class="_ _1"></span>hic<span class="_ _1"></span>h is one<span class="_ _1"></span> of t<span class="_ _1"></span>he mai<span class="_ _1"></span>n </div><div class="t m0 xe h5 yd ff1 fs3 fc0 sc0 lsc wsc">research content<span class="_ _1"></span>s in <span class="_ _1"></span>machine vi<span class="_ _1"></span>sion. For t<span class="_ _1"></span>he <span class="_ _1"></span>multi-scale and </div><div class="t m0 xe h5 ye ff1 fs3 fc0 sc0 lsd wsd">rotated image template matching, most of temp<span class="_ _0"></span>late matching<span class="_ _1"></span> </div><div class="t m0 xe h5 yf ff1 fs3 fc0 sc0 lse wse">algorithms usually form<span class="_ _0"></span> a templates collection w<span class="_ _1"></span>ith different </div><div class="t m0 xe h5 y10 ff1 fs3 fc0 sc0 ls9 wsf">scaling ratios templates, and then the tem<span class="_ _0"></span>plates in the </div><div class="t m0 xe h5 y11 ff1 fs3 fc0 sc0 lsf ws10">collection are <span class="_ _1"></span>matched se<span class="_ _1"></span>parately<span class="_ _1"></span>. The algorithm <span class="_ _1"></span>w<span class="_ _1"></span>ill greatly </div><div class="t m0 xe h5 y12 ff1 fs3 fc0 sc0 ls10 ws11">inc<span class="_ _0"></span>rea<span class="_ _0"></span>se th<span class="_ _0"></span>e ca<span class="_ _0"></span>lcul<span class="_ _0"></span>ati<span class="_ _0"></span>on bu<span class="_ _0"></span>rden of<span class="_ _0"></span> tem<span class="_ _0"></span>pla<span class="_ _0"></span>te m<span class="_ _0"></span>atch<span class="_ _0"></span>ing,<span class="_ _0"></span> and<span class="_ _0"></span> the </div><div class="t m0 xe h5 y13 ff1 fs3 fc0 sc0 lsa ws12">matching efficiency will be greatly reduced. This paper </div><div class="t m0 xe h5 y14 ff1 fs3 fc0 sc0 ls11 ws13">propos<span class="_ _1"></span>es an al<span class="_ _1"></span>go<span class="_ _1"></span>rith<span class="_ _1"></span>m for mul<span class="_ _1"></span>ti-<span class="_ _1"></span>scal<span class="_ _1"></span>e and r<span class="_ _1"></span>otat<span class="_ _1"></span>ed i<span class="_ _1"></span>mage </div><div class="t m0 xe h5 y15 ff1 fs3 fc0 sc0 lsa ws14">template matching. The algorithm first co<span class="_ _1"></span>mputes the ring </div><div class="t m0 xe h5 y16 ff1 fs3 fc0 sc0 ls12 ws15">projection vector of the te<span class="_ _1"></span>mplate<span class="_ _1"></span>, and then, the ring projection </div><div class="t m0 xe h5 y17 ff1 fs3 fc0 sc0 ls0 ws16">of the scaled template can be obtained by ring projection </div><div class="t m0 xe h5 y18 ff1 fs3 fc0 sc0 ls13 ws17">vector con<span class="_ _1"></span>version. T<span class="_ _1"></span>he nor<span class="_ _1"></span>malized<span class="lsf ws18"> cros<span class="_ _1"></span>s correlati<span class="_ _1"></span>on is used to<span class="_ _1"></span> </span></div><div class="t m0 xe h5 y19 ff1 fs3 fc0 sc0 lsf ws19">calculate the si<span class="_ _1"></span>milarity between the new<span class="_ _1"></span> ring projection vector<span class="_ _1"></span> </div><div class="t m0 xe h5 y1a ff1 fs3 fc0 sc0 ls14 ws1a">and the ring projecti<span class="_ _1"></span>on vector of<span class="_ _1"></span> each point of the scene image.<span class="_ _1"></span> </div><div class="t m0 xe h5 y1b ff1 fs3 fc0 sc0 ls15 ws1b">In the end t<span class="_ _1"></span>he similarities <span class="_ _1"></span>determine the <span class="_ _1"></span>opti<span class="_ _1"></span>mal matching </div><div class="t m0 xe h5 y1c ff1 fs3 fc0 sc0 ls16 ws1c">positi<span class="_ _1"></span>on a<span class="_ _1"></span>nd s<span class="_ _1"></span>cale r<span class="_ _1"></span>atio.<span class="_ _1"></span> Ex<span class="_ _1"></span>peri<span class="_ _1"></span>mental r<span class="_ _1"></span>es<span class="_ _1"></span>ults s<span class="_ _1"></span>how<span class="_ _1"></span> that t<span class="_ _1"></span>he </div><div class="t m0 xe h5 y1d ff1 fs3 fc0 sc0 ls14 ws1d">proposed alg<span class="_ _1"></span>orithm ca<span class="_ _1"></span>n accurately<span class="_ _1"></span> find the corre<span class="_ _1"></span>ct matchin<span class="_ _1"></span>g </div><div class="t m0 xe h5 y1e ff1 fs3 fc0 sc0 ls17 ws1e">po<span class="_ _0"></span>si<span class="_ _0"></span>tion<span class="_ _0"></span> for<span class="_ _0"></span> mu<span class="_ _0"></span>lti<span class="_ _0"></span>-sc<span class="_ _0"></span>al<span class="_ _0"></span>e and<span class="_ _0"></span> ro<span class="_ _0"></span>tat<span class="_ _0"></span>ed im<span class="_ _0"></span>ag<span class="_ _0"></span>e tem<span class="_ _0"></span>p<span class="_ _0"></span>late<span class="_ _0"></span> ma<span class="_ _0"></span>tch<span class="_ _0"></span>in<span class="_ _0"></span>g. </div><div class="t m0 xf h6 y1f ff3 fs3 fc0 sc0 ls18 ws1f">Keywords-temp<span class="_ _1"></span>late matching;<span class="_ _1"></span> ring <span class="_ _1"></span>projection <span class="_ _1"></span>vector; <span class="_ _1"></span>vector </div><div class="t m0 xe h6 y20 ff3 fs3 fc0 sc0 ls19 ws20">conve<span class="_ _0"></span>rsion; mul<span class="_ _0"></span>ti-scal<span class="_ _0"></span>e and rotat<span class="_ _0"></span>ed </div><div class="t m0 x10 h3 y21 ff2 fs1 fc0 sc0 ls1a ws0">I.<span class="ff5 ls0"> <span class="_ _2"> </span><span class="ff2 fs4"> <span class="fs1">I</span><span class="ls1b">NTRODUCTION <span class="_ _1"></span></span><span class="fs1"> </span></span></span></div><div class="t m0 xf h3 y22 ff2 fs1 fc0 sc0 ls1c ws21">Template m<span class="_ _0"></span>atching<span class="_ _0"></span> is one of<span class="_ _0"></span> the main<span class="_ _0"></span> research c<span class="_ _0"></span>ontents<span class="_ _0"></span> </div><div class="t m0 xe h3 y23 ff2 fs1 fc0 sc0 ls1d ws22">in graph<span class="_ _0"></span>ic im<span class="_ _0"></span>age processi<span class="_ _0"></span>ng, w<span class="_ _0"></span>hich has im<span class="_ _0"></span>portant<span class="_ _0"></span> </div><div class="t m0 xe h3 y24 ff2 fs1 fc0 sc0 ls1e ws23">applica<span class="_ _0"></span>tions in m<span class="_ _0"></span>achin<span class="_ _0"></span>e visio<span class="_ _0"></span>n, targe<span class="_ _0"></span>t re<span class="_ _0"></span>cognition<span class="_ _0"></span>, m<span class="_ _0"></span>edical<span class="_ _0"></span> </div><div class="t m0 xe h3 y25 ff2 fs1 fc0 sc0 ls1f ws24">image processin<span class="_ _0"></span>g and othe<span class="_ _0"></span>r fields<span class="_ _0"></span> [1-3]. </div><div class="t m0 xf h3 y26 ff2 fs1 fc0 sc0 ls20 ws25">Template m<span class="_ _0"></span>atching is a<span class="_ _0"></span> process <span class="_ _0"></span>of searchin<span class="_ _0"></span>g a small<span class="_ _0"></span> </div><div class="t m0 xe h3 y27 ff2 fs1 fc0 sc0 ls21 ws26">template<span class="_ _0"></span> image in <span class="_ _0"></span>the s<span class="_ _0"></span>cene im<span class="_ _0"></span>age. The gene<span class="_ _0"></span>ral tem<span class="_ _0"></span>plate </div><div class="t m0 xe h3 y28 ff2 fs1 fc0 sc0 ls22 ws27">matchin<span class="_ _0"></span>g algorith<span class="_ _0"></span>ms usu<span class="_ _0"></span>ally use the m<span class="_ _0"></span>ethod of pix<span class="_ _0"></span>el-by-</div><div class="t m0 xe h3 y29 ff2 fs1 fc0 sc0 ls23 ws28">point search t<span class="_ _0"></span>o compare th<span class="_ _0"></span>e sim<span class="_ _0"></span>ilarity<span class="_ _0"></span> between th<span class="_ _0"></span>e template<span class="_ _0"></span> </div><div class="t m0 xe h3 y2a ff2 fs1 fc0 sc0 ls23 ws29">and the su<span class="_ _0"></span>b-images <span class="_ _0"></span>of ea<span class="_ _0"></span>ch <span class="ls24 ws2a">pos<span class="_ _0"></span>ition<span class="_ _0"></span> in <span class="_ _0"></span>the scene <span class="_ _0"></span>images,<span class="_ _0"></span> </span></div><div class="t m0 xe h3 y2b ff2 fs1 fc0 sc0 ls21 ws2b">when the s<span class="_ _0"></span>im<span class="_ _0"></span>ilarity is<span class="_ _0"></span> the m<span class="_ _0"></span>aximum, th<span class="_ _0"></span>e positi<span class="_ _0"></span>on is <span class="_ _0"></span>the target<span class="_ _0"></span> </div><div class="t m0 xe h3 y2c ff2 fs1 fc0 sc0 ls25 ws2c">location.<span class="_ _0"></span> The com<span class="_ _0"></span>putation<span class="_ _0"></span>al ti<span class="_ _0"></span>me is the k<span class="_ _0"></span>ey t<span class="_ _0"></span>o evaluation<span class="_ _0"></span> th<span class="_ _0"></span>e </div><div class="t m0 xe h3 y2d ff2 fs1 fc0 sc0 ls26 ws2d">algorithm<span class="_ _0"></span>, so <span class="_ _0"></span>a variety<span class="_ _0"></span> of <span class="_ _0"></span>template<span class="_ _0"></span> matchin<span class="_ _0"></span>g algor<span class="_ _0"></span>ithm<span class="_ _0"></span>s have </div><div class="t m0 xe h3 y2e ff2 fs1 fc0 sc0 ls1e ws2e">been pro<span class="_ _0"></span>posed [4-6] to<span class="_ _0"></span> reduce<span class="_ _0"></span> the computati<span class="_ _0"></span>onal<span class="_ _0"></span> burden..<span class="_ _0"></span> </div><div class="t m0 xf h3 y2f ff2 fs1 fc0 sc0 ls1f ws2f">The above<span class="_ _0"></span> algo<span class="_ _0"></span>rithm los<span class="_ _0"></span>es its f<span class="_ _0"></span>un<span class="_ _0"></span>ction w<span class="_ _0"></span>hen the di<span class="_ _0"></span>rection<span class="_ _0"></span> </div><div class="t m0 xe h3 y30 ff2 fs1 fc0 sc0 ls27 ws30">or size <span class="_ _0"></span>of th<span class="_ _0"></span>e objects <span class="_ _0"></span>are dif<span class="_ _0"></span>ferent f<span class="_ _0"></span>rom th<span class="_ _0"></span>e templat<span class="_ _0"></span>e. To </div><div class="t m0 xe h3 y31 ff2 fs1 fc0 sc0 ls26 ws31">solve th<span class="_ _0"></span>is pr<span class="_ _0"></span>oblem, th<span class="_ _0"></span>e init<span class="_ _0"></span>ial app<span class="_ _0"></span>roach is to c<span class="_ _0"></span>reate <span class="_ _0"></span>a seri<span class="_ _0"></span>es of<span class="_ _0"></span> </div><div class="t m0 xe h3 y32 ff2 fs1 fc0 sc0 ls27 ws32">templates<span class="_ _0"></span> [7] in <span class="_ _0"></span>diffe<span class="_ _0"></span>rent dire<span class="_ _0"></span>ctions an<span class="_ _0"></span>d sizes<span class="_ _0"></span>, and th<span class="_ _0"></span>en<span class="_ _0"></span> </div><div class="t m0 xe h3 y33 ff2 fs1 fc0 sc0 ls5 ws33">partic<span class="_ _0"></span>ipate in the cal<span class="_ _0"></span>culation w<span class="_ _0"></span>ith each template. Obvi<span class="_ _0"></span>ously<span class="_ _0"></span>, </div><div class="t m0 xe h3 y34 ff2 fs1 fc0 sc0 ls1c ws34">this m<span class="_ _0"></span>ethod is too m<span class="_ _0"></span>uch com<span class="_ _0"></span>putation to use <span class="_ _0"></span>in the actu<span class="_ _0"></span>al </div><div class="t m0 xe h3 y35 ff2 fs1 fc0 sc0 ls28 ws35">system<span class="_ _0"></span>. The <span class="_ _1"></span>matching<span class="_ _0"></span> method bas<span class="_ _0"></span>ed on invariant<span class="_ _0"></span> moment [8] </div><div class="t m0 xe h3 y36 ff2 fs1 fc0 sc0 ls27 ws36">can be u<span class="_ _0"></span>sed to<span class="_ _0"></span> solve the<span class="_ _0"></span> rotati<span class="_ _0"></span>on problem<span class="_ _0"></span> of the <span class="_ _0"></span>target<span class="_ _0"></span>, but i<span class="_ _0"></span>t </div><div class="t m0 x11 h3 y37 ff6 fs1 fc0 sc0 ls1c ws37">can&#8217;t fin<span class="_ _0"></span>d th<span class="_ _0"></span>e multi<span class="ff2 ls28 ws38">-s<span class="_ _0"></span>cale im<span class="_ _0"></span>age tem<span class="_ _0"></span>plate, and it is<span class="_ _0"></span> very<span class="_ _0"></span> </span></div><div class="t m0 x11 h3 y38 ff2 fs1 fc0 sc0 ls29 ws39">sensitive t<span class="_ _0"></span>o the no<span class="_ _0"></span>ise. At <span class="_ _0"></span>present,<span class="_ _0"></span> most of<span class="_ _0"></span> the tem<span class="_ _0"></span>plate </div><div class="t m0 x11 h3 y39 ff2 fs1 fc0 sc0 ls21 ws3a">matchin<span class="_ _0"></span>g algorithm<span class="_ _0"></span>s do n<span class="_ _0"></span>ot solve th<span class="_ _0"></span>e probl<span class="_ _0"></span>em of mu<span class="_ _0"></span>lti-scale<span class="_ _0"></span> </div><div class="t m0 x11 h3 y3a ff2 fs1 fc0 sc0 ls0 ws3b">image tem<span class="_ _0"></span>plate m<span class="_ _0"></span>atching w<span class="_ _0"></span>ell. </div><div class="t m0 x12 h3 y3b ff2 fs1 fc0 sc0 ls27 ws3c"> In<span class="_ _0"></span> this pape<span class="_ _0"></span>r, we propose a te<span class="_ _0"></span>mplate matchin<span class="_ _0"></span>g for multi-</div><div class="t m0 x11 h3 y3c ff2 fs1 fc0 sc0 ls23 ws3d">scale and<span class="_ _0"></span> rotate<span class="_ _0"></span>d im<span class="_ _0"></span>age templa<span class="_ _0"></span>te m<span class="_ _0"></span>atching. At th<span class="_ _0"></span>e begin<span class="_ _0"></span>ning,<span class="_ _0"></span> </div><div class="t m0 x11 h3 y3d ff2 fs1 fc0 sc0 ls1f ws31">the tem<span class="_ _0"></span>plate an<span class="_ _0"></span>d scene<span class="_ _0"></span> image <span class="_ _0"></span>are re<span class="_ _0"></span>duced in <span class="_ _0"></span>resolut<span class="_ _0"></span>ion. The<span class="_ _0"></span>n </div><div class="t m0 x11 h3 y3e ff2 fs1 fc0 sc0 ls27 ws3e">template<span class="_ _0"></span> vect<span class="_ _0"></span>or c<span class="_ _0"></span>an be <span class="_ _0"></span>obta<span class="_ _0"></span>ined by<span class="_ _0"></span> calcu<span class="_ _0"></span>lating th<span class="_ _0"></span>e rin<span class="_ _0"></span>g </div><div class="t m0 x11 h3 y3f ff2 fs1 fc0 sc0 ls2a ws3f">projec<span class="_ _0"></span>tion ve<span class="_ _0"></span>ctor <span class="_ _0"></span>of the<span class="_ _0"></span> tem<span class="_ _0"></span>plate. Afte<span class="_ _0"></span>r that,<span class="_ _0"></span> by<span class="_ _0"></span> transform<span class="_ _0"></span>ing </div><div class="t m0 x11 h3 y40 ff2 fs1 fc0 sc0 ls2b ws40">the ring proj<span class="_ _1"></span>ect<span class="_ _1"></span>ion vector<span class="_ _1"></span>, we ca<span class="_ _1"></span>n obtain the ri<span class="_ _1"></span>ng projec<span class="_ _1"></span>tion </div><div class="t m0 x11 h3 y41 ff2 fs1 fc0 sc0 ls1d ws41">vector<span class="_ _0"></span> of<span class="_ _0"></span> scaled t<span class="_ _0"></span>emplat<span class="_ _0"></span>e. In<span class="_ _0"></span> the <span class="_ _0"></span>end, calcul<span class="_ _0"></span>ate<span class="_ _0"></span> the sim<span class="_ _0"></span>ilarit<span class="_ _0"></span>y </div><div class="t m0 x11 h3 y42 ff2 fs1 fc0 sc0 ls1f ws42">of the transform<span class="_ _0"></span>ed ring pr<span class="_ _0"></span>ojection vect<span class="_ _0"></span>or with the ring<span class="_ _0"></span> </div><div class="t m0 x11 h3 y43 ff2 fs1 fc0 sc0 ls2c ws43">projec<span class="_ _0"></span>tion ve<span class="_ _0"></span>ctor of<span class="_ _0"></span> each pos<span class="_ _0"></span><span class="ls25 ws44">ition<span class="_ _0"></span> on the scene image, w<span class="_ _0"></span>e </span></div><div class="t m0 x11 h3 y44 ff2 fs1 fc0 sc0 ls1d ws45">can dete<span class="_ _0"></span>rmine the o<span class="_ _0"></span>ptimal mat<span class="_ _0"></span>ching point<span class="_ _0"></span>. </div><div class="t m0 x13 h3 y45 ff2 fs1 fc0 sc0 ls1a ws0">II.<span class="ff5 ls0"> <span class="_ _3"> </span><span class="ff2">T</span></span></div><div class="t m0 x14 h3 y46 ff2 fs4 fc0 sc0 ls2d ws0">EMPLA<span class="_ _1"></span>TE <span class="_ _1"></span><span class="fs1 ls0">M</span><span class="ls1b">ATCHING <span class="_ _1"></span><span class="fs1 ls0">A</span><span class="ls2e">LGORITHM<span class="fs1 ls0"> </span></span></span></div><div class="t m0 x12 h3 y47 ff2 fs1 fc0 sc0 ls24 ws46">In or<span class="_ _0"></span>der t<span class="_ _0"></span>o solve th<span class="_ _0"></span>e change<span class="_ _0"></span>s of r<span class="_ _0"></span>otation f<span class="_ _0"></span>or th<span class="_ _0"></span>e scene<span class="_ _0"></span> </div><div class="t m0 x11 h3 y48 ff2 fs1 fc0 sc0 ls1d ws44">image in<span class="_ _0"></span> the tem<span class="_ _0"></span>plate m<span class="_ _0"></span>atching, the rin<span class="_ _0"></span>g pro<span class="_ _0"></span>jection <span class="_ _0"></span>of the<span class="_ _0"></span> </div><div class="t m0 x11 h3 y49 ff2 fs1 fc0 sc0 ls26 ws47">template<span class="_ _0"></span> and s<span class="_ _0"></span>cene im<span class="_ _0"></span>age are sel<span class="_ _0"></span>ected as<span class="_ _0"></span> the m<span class="_ _0"></span>atching </div><div class="t m0 x11 h3 y4a ff2 fs1 fc0 sc0 ls2f ws0">feat<span class="_ _1"></span>ure. </div><div class="t m0 x11 h7 y4b ff7 fs1 fc0 sc0 ls30 ws0">A.<span class="ff8 ls0"> <span class="_"> </span></span><span class="ls2a ws48">Ring Projectio<span class="_ _1"></span>n Transformation </span></div><div class="t m0 x12 h3 y4c ff2 fs1 fc0 sc0 ls31 ws49">Ring proje<span class="_ _1"></span>ction transfor<span class="_ _1"></span>mation proce<span class="_ _1"></span>ss [9] was propo<span class="_ _1"></span>sed </div><div class="t m0 x11 h3 y4d ff2 fs1 fc0 sc0 ls6 ws4a">for redu<span class="_ _0"></span>cing the compu<span class="_ _0"></span>tation<span class="_ _0"></span>al burden.<span class="_ _0"></span> In order to com<span class="_ _0"></span>pute </div><div class="t m0 x11 h3 y4e ff2 fs1 fc0 sc0 ls1f ws4b">the rin<span class="_ _0"></span>g pr<span class="_ _0"></span>ojection<span class="_ _0"></span> vect<span class="_ _0"></span>or of<span class="_ _0"></span> the t<span class="_ _0"></span>emplate<span class="_ _0"></span> image, w<span class="_ _0"></span>e nee<span class="_ _0"></span>d to<span class="_ _0"></span> </div><div class="t m0 x11 h3 y4f ff2 fs1 fc0 sc0 ls1c ws4c">turn the s<span class="_ _0"></span>quare<span class="_ _0"></span> image into<span class="_ _0"></span> a circl<span class="_ _0"></span>e, as show<span class="_ _0"></span>n in<span class="_ _0"></span> Fig. 1, f </div><div class="t m0 x11 h3 y50 ff2 fs1 fc0 sc0 ls26 ws4d">represents<span class="_ _0"></span> the t<span class="_ _0"></span>emplate<span class="_ _0"></span> image.<span class="_ _0"></span> </div><div class="t m0 x15 h3 y51 ff2 fs1 fc0 sc0 ls0 ws0"> </div><div class="t m0 x14 h8 y52 ff2 fs4 fc0 sc0 ls32 ws4e">Fig<span class="_ _1"></span>ur<span class="_ _1"></span>e 1<span class="_ _1"></span>.<span class="_ _1"></span> <span class="ff5 ls0 ws0"> <span class="_ _4"> </span></span><span class="ls33 ws4f">The<span class="_ _0"></span> ring pro<span class="_ _0"></span>jectio<span class="_ _0"></span>n. </span></div><div class="t m0 x12 h3 y53 ff2 fs1 fc0 sc0 ls34 ws50">The p<span class="_ _0"></span>rojecti<span class="_ _0"></span>on v<span class="_ _0"></span>ector<span class="_ _0"></span> is c<span class="_ _0"></span>ompos<span class="_ _0"></span>ed <span class="_ _0"></span>of th<span class="_ _0"></span>e pix<span class="_ _0"></span>el m<span class="_ _0"></span>ean on<span class="_ _0"></span> </div><div class="t m0 x11 h3 y54 ff2 fs1 fc0 sc0 ls35 ws51">the circum<span class="_ _0"></span>ference of conc<span class="_ _0"></span>entri<span class="_ _0"></span>c circles of<span class="_ _0"></span> different ra<span class="_ _0"></span>dii, </div><div class="t m0 x11 h3 y55 ff2 fs1 fc0 sc0 ls27 ws52">where <span class="_ _0"></span>the ring<span class="_ _0"></span> pr<span class="_ _0"></span>ojection<span class="_ _0"></span> value <span class="_ _0"></span>P (r) <span class="_ _0"></span>of a <span class="_ _0"></span>circle<span class="_ _0"></span> with<span class="_ _0"></span> a radiu<span class="_ _0"></span>s </div><div class="t m0 x11 h3 y56 ff2 fs1 fc0 sc0 ls36 ws53">of r is:<span class="_ _0"></span> </div><div class="t m0 x16 h9 y57 ff9 fs3 fc0 sc0 ls0 ws0">45</div><div class="t m0 x17 ha y58 ff9 fs5 fc0 sc0 ls37 ws54">2018 3rd IEEE International Conference on Image, Vision and Computing</div><div class="t m0 xe hb y59 ff2 fs3 fc0 sc0 ls38 ws55">978-1-5386-4991-6/18/$31.00 &#169;2018 IEEE</div></div><div class="pi" data-data='{"ctm":[1.568627,0.000000,0.000000,1.568627,0.000000,0.000000]}'></div></div> </body> </html>
评论
    相关推荐
    • TemplateMatching.zip
      计算机视觉模板匹配,电子书,对于模式识别特别是模板匹配进行识别,这是十分有用的书.
    • NCC Template matching.zip
      NCC模板匹配,可以直接运行,加载本地图片。
    • Template matching.zip
      基于opcv的模板匹配,支持单对单和多对多。
    • Template-matching.rar
      在图像中,根据模板匹配,搜索到所需要的测试图像对应项目
    • matlab开发-TemplateMatching
      matlab开发-TemplateMatching。使用性能索引方法进行模板匹配
    • Template-matching:使用归一化互相关的模板匹配
      模板匹配 使用归一化互相关的模板匹配 优点: 易于实现。 如果模板/数据库很小,则可以快速工作。 缺点: 如果模板/数据库较大,则速度较慢。 对任何类型的旋转或比例变化都非常敏感。 这是最终输出的屏幕截图...
    • Template Matching.rar
      python-OpenCV实现边缘模板匹配算法。介绍一种新的模板匹配算法,主要是基于图像边缘梯度,它对图像光照与像素迁移都有很强的抗干扰能力
    • TemplateMatching.zip
      matlab开发-TemplateMatching模板匹配
    • TemplateMatching - 副本.rar
      使用鼠标画框选择目标,使用模板匹配进行视觉追踪
    • codesforimageprocessing.rar
      实现简单图像处理,包括256色转灰度图、Hough变换、Walsh变换、中值滤波、二值化变换、亮度增减、傅立叶变换、反色、取对数、取指数、图像平移、图像旋转、图像细化、图像缩放、图像镜像、均值滤波、对比度拉伸、拉普拉斯锐化(边缘检测)、方块编码、梯度锐化、灰度均衡、直方图均衡、离散余弦变换、维纳滤波处理、逆滤波处理、阈值变换、高斯平滑。