BP神经网络算法完整学习资料图象识别技术入门必备_C#_C#编程

<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/62781aead7996104df5759ff/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/62781aead7996104df5759ff/bg1.jpg"><div class="c x0 y1 w2 h2"><div class="t m0 x1 h3 y2 ff1 fs0 fc0 sc0 ls0 ws0">第<span class="ff2">1<span class="_ _0"></span>1<span class="ff1">章</span> <span class="ff1">数字识别系统</span></span></div><div class="t m0 x2 h4 y3 ff2 fs1 fc0 sc0 ls0 ws0">1<span class="_ _1"></span>1.1 <span class="ff1">系统简介</span></div><div class="t m0 x3 h5 y4 ff1 fs2 fc0 sc0 ls0 ws0">该系统是为了辨认识别图像中的数字而设计的，它通过对图片的一系列处理，最后识别得</div><div class="t m0 x2 h5 y5 ff1 fs2 fc0 sc0 ls0 ws0">出图片中显示的数字。系统既可以单独使用，也可以把它作为一个识别系统的软件核心应用到</div><div class="t m0 x2 h5 y6 ff1 fs2 fc0 sc0 ls0 ws0">车牌识别等系统中去。</div><div class="t m0 x2 h4 y7 ff2 fs1 fc0 sc0 ls0 ws0">1<span class="_ _1"></span>1.2 <span class="ff1">系统基本技术要求</span></div><div class="t m0 x3 h5 y8 ff1 fs2 fc0 sc0 ls0 ws0">下面是系统具体要达到的基本技术要求</div><div class="t m0 x3 h5 y9 ff1 fs2 fc0 sc0 ls0 ws0">（<span class="ff3">1</span>）输入图片中可以含有多个数字；</div><div class="t m0 x3 h5 ya ff1 fs2 fc0 sc0 ls0 ws0">（<span class="ff3">2</span>）数字的识别准确率大于<span class="ff3">90</span>％；</div><div class="t m0 x3 h5 yb ff1 fs2 fc0 sc0 ls0 ws0">（<span class="ff3">3</span>）每张图片的处理时间（识别时间）不能大于<span class="ff3">1</span>秒；</div><div class="t m0 x3 h5 yc ff1 fs2 fc0 sc0 ls0 ws0">（<span class="ff3">4</span>）对图片噪声具有较强的适应性；</div><div class="t m0 x3 h5 yd ff1 fs2 fc0 sc0 ls0 ws0">（<span class="ff3">5</span>）系统要能长时间无故障运行；</div><div class="t m0 x3 h5 ye ff1 fs2 fc0 sc0 ls0 ws0">（<span class="ff3">6</span>）系统的操作要求简单。</div><div class="t m0 x2 h4 yf ff2 fs1 fc0 sc0 ls0 ws0">1<span class="_ _1"></span>1.3 <span class="ff1">系统中用到的关键技术</span></div><div class="t m0 x3 h5 y10 ff1 fs2 fc0 sc0 ls0 ws0">在本系统中用到了好多图像处理中的相关技术：比如灰度化、二值化、图像内容自动调整</div><div class="t m0 x2 h5 y11 ff1 fs2 fc0 sc0 ls0 ws0">去离散点、图像的缩放、细化、曲线平滑、曲线去枝桠操作等，最后还使用了神经网络对提取</div><div class="t m0 x2 h5 y12 ff1 fs2 fc0 sc0 ls0 ws0">到的数字信息进行分析判断。</div><div class="t m0 x2 h4 y13 ff2 fs1 fc0 sc0 ls0 ws0">1<span class="_ _1"></span>1.4 <span class="ff1">系统软硬件平台</span></div><div class="t m0 x2 h6 y14 ff2 fs3 fc0 sc0 ls0 ws0">1<span class="_ _1"></span>1.4.1 <span class="ff1">系统的硬件平台</span></div><div class="t m0 x3 h5 y15 ff1 fs2 fc0 sc0 ls0 ws0">因为系统运行的过程当中，主要进行的都是图像处理，在这个过程当中要进行大量的数据</div><div class="t m0 x2 h5 y16 ff1 fs2 fc0 sc0 ls0 ws0">处理，所以处理器和内存要求比较高，<span class="ff3">CPU </span>要求主频在<span class="ff3">600hz </span>以上（含<span class="ff3">600hz</span>），内存在<span class="_ _2"> </span><span class="ff3">128</span></div><div class="t m0 x2 h5 y17 ff1 fs2 fc0 sc0 ls0 ws0">兆以上（含<span class="ff3">128 </span>兆）。</div></div></div><div class="pi" data-data='{"ctm":[1.839080,0.000000,0.000000,1.839080,0.000000,0.000000]}'></div></div> </body> </html>

<div id="pf2" class="pf w0 h0" data-page-no="2"><div class="pc pc2 w0 h0"><img class="bi x0 y0 w1 h1" alt="" src="https://static.pudn.com/prod/directory_preview_static/62781aead7996104df5759ff/bg2.jpg"><div class="c x0 y1 w2 h2"><div class="t m0 x2 h4 y18 ff2 fs3 fc0 sc0 ls0 ws0">1<span class="_ _1"></span>1.4<span class="fs1">.</span>2 <span class="ff1 fs1">系统的软件平台</span></div><div class="t m0 x3 h5 y19 ff1 fs2 fc0 sc0 ls0 ws0">系统可以运行于任何一台装有<span class="ff3">Windows98 </span>、<span class="ff3">Windows2000 </span>或者<span class="ff3">Windows XP<span class="_ _3"></span> <span class="ff1">的机子上。程</span></span></div><div class="t m0 x2 h5 y1a ff1 fs2 fc0 sc0 ls0 ws0">序调试时，需要使用<span class="ff3">Microsoft V<span class="_ _1"></span>isual C++ 6.0<span class="_ _4"></span><span class="ff1">（</span>SP6<span class="ff1">）。</span></span></div><div class="t m0 x2 h4 y1b ff2 fs3 fc0 sc0 ls0 ws0">1<span class="_ _1"></span>1.4<span class="fs1">.</span>3 <span class="ff1 fs1">本章源码使用说明</span></div><div class="t m0 x3 h5 y1c ff1 fs2 fc0 sc0 ls0 ws0">本系统的全部源码在本书配套光盘的【<span class="ff3">\Chap1<span class="_ _3"></span>1\<span class="ff1">】目录下。可参看</span>readme.txt <span class="ff1">进行使用。</span></span></div><div class="t m0 x2 h4 y1d ff2 fs1 fc0 sc0 ls0 ws0">1<span class="_ _1"></span>1.5 <span class="ff1">系统实现</span></div><div class="t m0 x3 h5 y1e ff1 fs2 fc0 sc0 ls0 ws0">系统在实现的过程当中，先分解成两个大块，就是图像预处理模块和数字识别模块。其中</div><div class="t m0 x2 h5 y1f ff1 fs2 fc0 sc0 ls0 ws0">图像像预处理块在对图像进行了一系列变换后把最后提取到的数字字符提交给数字识别模块，</div><div class="t m0 x2 h5 y20 ff1 fs2 fc0 sc0 ls0 ws0">然后进行识别并给处结果。在这里用到了很多先进的图像预处理技术及神经网络技术。</div><div class="t m0 x2 h4 y21 ff2 fs3 fc0 sc0 ls0 ws0">1<span class="_ _1"></span>1.5<span class="fs1">.</span>1 <span class="ff1 fs1">系统流程图</span></div><div class="t m0 x3 h5 y22 ff1 fs2 fc0 sc0 ls0 ws0">本系统总的流程结构如图<span class="ff3">1<span class="_ _3"></span>1-1<span class="ff1">所示。</span></span></div><div class="t m0 x4 h7 y23 ff1 fs4 fc0 sc0 ls0 ws0">图<span class="ff2">1<span class="_ _3"></span>1-1 <span class="ff1">系统总流程图</span></span></div><div class="t m0 x3 h5 y24 ff1 fs2 fc0 sc0 ls0 ws0">其中图像预处理的流程如图<span class="ff3">1<span class="_ _3"></span>1-2 <span class="ff1">所示。</span></span></div></div></div><div class="pi" data-data='{"ctm":[1.839080,0.000000,0.000000,1.839080,0.000000,0.000000]}'></div></div>

<div id="pf3" class="pf w0 h0" data-page-no="3"><div class="pc pc3 w0 h0"><img class="bi x0 y0 w1 h1" alt="" src="https://static.pudn.com/prod/directory_preview_static/62781aead7996104df5759ff/bg3.jpg"><div class="c x0 y1 w2 h2"><div class="t m0 x5 h7 y25 ff1 fs4 fc0 sc0 ls0 ws0">图<span class="ff2">1<span class="_ _3"></span>1-2 <span class="ff1">图像预处理流程图</span></span></div><div class="t m0 x3 h5 y26 ff1 fs2 fc0 sc0 ls0 ws0">神经网络数字识别的具体流程如图<span class="ff3">1<span class="_ _3"></span>1-3 <span class="ff1">所示。</span></span></div><div class="t m0 x6 h7 y27 ff1 fs4 fc0 sc0 ls0 ws0">图<span class="ff2">1<span class="_ _3"></span>1-3 <span class="ff1">神经网络数字识别流程图</span></span></div><div class="t m0 x2 h6 y28 ff2 fs1 fc0 sc0 ls0 ws0">1<span class="_ _1"></span>1.5.2 <span class="ff1 fs3">程序实现</span></div><div class="t m0 x3 h5 y29 ff1 fs2 fc0 sc0 ls0 ws0">整个系统的程序实现分为图像预处理和神经网络识别两大模块。在图像预处理的过程当中</div><div class="t m0 x2 h5 y2a ff1 fs2 fc0 sc0 ls0 ws0">我们采用了许多图像处理的技术，最后把每个数字的特征提取出来。这些技术包括图像数据读</div><div class="t m0 x2 h5 y2b ff1 fs2 fc0 sc0 ls0 ws0">取、图像的灰度化、二值化、图像的调整、离散噪声点的去除、字符的切分、图像的缩放、字</div><div class="t m0 x2 h5 y2c ff1 fs2 fc0 sc0 ls0 ws0">符的细化、字符的平滑、图像的求梯度等图像处理技术，最后是数字字符特征的提取。其结果</div><div class="t m0 x2 h5 y2d ff1 fs2 fc0 sc0 ls0 ws0">再利用神经网络（这里我们选用<span class="ff3">BP</span>网络）进行字符识别。</div><div class="t m0 x3 h5 y2e ff1 fs2 fc0 sc0 ls0 ws0">利用神经网络进行字符识别的过程主要包括网络的训练、数据的读取、字符的判定、结果</div><div class="t m0 x2 h5 y2f ff1 fs2 fc0 sc0 ls0 ws0">的输出。下面按照程序执行的顺序介绍整个程序并逐一分析每一步的关键代码。最后简要讲述</div><div class="t m0 x2 h8 y30 ff1 fs2 fc0 sc0 ls0 ws0">一下程序的使用方法、<span class="ff4">注意事项</span>以及本章<span class="ff4">小</span>结。</div><div class="t m0 x2 h8 y31 ff2 fs2 fc0 sc0 ls0 ws0">1. <span class="ff1">图像数据的读取、<span class="ff4">保</span>存<span class="ff4">与屏幕</span>显示等基本<span class="ff4">函</span>数</span></div><div class="t m0 x3 h8 y32 ff1 fs2 fc0 sc0 ls0 ws0">要进行图像分析和处理<span class="ff4">首</span>先就要得到图像的数据，这些数据包括图像的<span class="ff4">宽</span>、高、每个<span class="ff4">象素</span></div><div class="t m0 x2 h8 y33 ff1 fs2 fc0 sc0 ls0 ws0">点的<span class="ff4">颜色</span>值。因为每<span class="ff4">种文</span>件都有它自<span class="ff4">己</span>的存放<span class="ff4">格式</span>，下面简单介绍<span class="ff3">BMP<span class="_ _3"></span> <span class="ff4">文<span class="ff1">件的存放</span>格式<span class="ff1">。本</span></span></span></div><div class="t m0 x2 h8 y34 ff1 fs2 fc0 sc0 ls0 ws0">程序采用的输入图像<span class="ff4">即</span>为<span class="ff3">bmp<span class="ff4">位</span></span>图。</div></div></div><div class="pi" data-data='{"ctm":[1.839080,0.000000,0.000000,1.839080,0.000000,0.000000]}'></div></div>

<div id="pf4" class="pf w0 h0" data-page-no="4"><div class="pc pc4 w0 h0"><img class="bi x0 y0 w1 h1" alt="" src="https://static.pudn.com/prod/directory_preview_static/62781aead7996104df5759ff/bg4.jpg"><div class="c x0 y1 w2 h2"><div class="t m0 x3 h8 y35 ff3 fs2 fc0 sc0 ls0 ws0">BMP<span class="ff1">（</span>bitmap <span class="ff1">的缩<span class="ff4">写</span>）<span class="ff4">文</span>件<span class="ff4">格式</span>是</span>W<span class="_ _3"></span>indows <span class="ff1">本<span class="ff4">身</span>可以<span class="ff4">直接</span>提<span class="ff4">供</span>读取<span class="ff4">支持</span>的<span class="ff4">位</span>图<span class="ff4">文</span>件<span class="ff4">格式</span></span></div><div class="t m0 x2 h8 y36 ff1 fs2 fc0 sc0 ls0 ws0">一个<span class="ff3">BMP <span class="ff4">格式</span></span>的<span class="ff4">文</span>件通<span class="ff4">常</span>有<span class="ff3">.BMP<span class="_ _3"></span> <span class="ff1">的<span class="ff4">扩展名</span>，<span class="ff4">但</span>有一些是以</span>.rle<span class="ff1">为<span class="ff4">扩展名</span>的，</span>rle <span class="ff1">的<span class="ff4">意思</span>是行程</span></span></div><div class="t m0 x2 h8 y37 ff1 fs2 fc0 sc0 ls0 ws0">长度<span class="ff4">编</span>码（<span class="ff3">run length encoding</span>）。这<span class="ff4">样</span>的<span class="ff4">文</span>件<span class="ff4">意味着</span>其使用的数据<span class="ff4">压</span>缩方法是<span class="_ _2"> </span><span class="ff3">BMP<span class="_ _3"></span> <span class="ff4">格式文<span class="ff1">件</span></span></span></div><div class="t m0 x2 h8 y38 ff4 fs2 fc0 sc0 ls0 ws0">支持<span class="ff1">的两</span>种<span class="ff3">rle <span class="ff1">方法中的一</span></span>种<span class="ff1">。<span class="ff3">BMP<span class="_ _3"></span> <span class="ff4">文<span class="ff1">件可用每</span>象素</span>1<span class="ff1">、</span>4<span class="ff1">、</span>8<span class="ff1">、</span>16 <span class="ff1">或</span>24 <span class="ff4">位<span class="ff1">来</span>编<span class="ff1">码</span>颜色<span class="ff1">信息，这</span></span></span></span></div><div class="t m0 x2 h8 y39 ff1 fs2 fc0 sc0 ls0 ws0">个<span class="ff4">位<span class="_ _4"></span></span>数<span class="ff4">称<span class="_ _4"></span></span>作图<span class="_ _4"></span>像的<span class="_ _4"></span><span class="ff4">颜色<span class="_ _4"></span>深</span>度<span class="_ _4"></span>，它<span class="_ _4"></span><span class="ff4">决</span>定<span class="_ _4"></span>了图<span class="_ _4"></span>像所<span class="_ _4"></span>含的最<span class="_ _4"></span>大<span class="_ _4"></span><span class="ff4">颜色<span class="_ _4"></span></span>数。<span class="_ _4"></span>一<span class="ff4">幅<span class="_ _5"></span><span class="ff3">1-<span class="_ _4"></span>bp<span class="_ _4"></span>p<span class="_ _3"></span><span class="ff1">（<span class="ff4">位<span class="_ _4"></span></span>每<span class="ff4">象<span class="_ _4"></span>素</span>，<span class="_ _4"></span><span class="ff3">b<span class="_ _4"></span>it</span></span></span></span></div><div class="t m0 x2 h8 y3a ff3 fs2 fc0 sc0 ls0 ws0">per pixel<span class="ff1">）的图像<span class="ff4">只</span>能有两<span class="ff4">种颜色</span>。而一<span class="ff4">幅</span></span>24-bpp <span class="ff1">的图像可以有<span class="ff4">超</span>过</span>16 <span class="ff1">兆<span class="ff4">种</span>不<span class="ff4">同</span>的<span class="ff4">颜色</span>。</span></div><div class="t m0 x3 h8 y3b ff1 fs2 fc0 sc0 ls0 ws0">一个<span class="ff4">典型<span class="ff3">BMP<span class="_ _3"></span> <span class="ff4">文<span class="ff1">件的结构。以</span></span>256 <span class="ff4">色<span class="ff1">也就是</span></span>8-bpp <span class="ff1">为<span class="ff4">例</span>，<span class="ff4">文</span>件<span class="ff4">被</span>分成<span class="ff4">四</span>个主要的部分：一</span></span></span></div><div class="t m0 x2 h8 y3c ff1 fs2 fc0 sc0 ls0 ws0">个<span class="ff4">位</span>图<span class="ff4">文</span>件<span class="ff4">头</span>，一个<span class="ff4">位</span>图信息<span class="ff4">头</span>，一个<span class="ff4">颜色表</span>（<span class="ff4">又称</span>为<span class="ff4">色表</span>）和<span class="ff4">位</span>图数据本<span class="ff4">身</span>。</div><div class="t m0 x3 h8 y3d ff4 fs2 fc0 sc1 ls0 ws0">位<span class="ff1">图</span>文<span class="ff1">件</span>头</div><div class="t m0 x3 h8 y3e ff4 fs2 fc0 sc0 ls0 ws0">位<span class="ff1">图</span>文<span class="ff1">件</span>头<span class="ff1">包含关于这个</span>文<span class="ff1">件的信息。如</span>从哪<span class="ff1">里</span>开始<span class="ff1">是</span>位<span class="ff1">图数据的定</span>位<span class="ff1">信息、</span>文<span class="ff1">件大</span>小<span class="ff1">等</span></div><div class="t m0 x2 h8 y3f ff1 fs2 fc0 sc0 ls0 ws0">等。以下是<span class="ff4">位</span>图<span class="ff4">文</span>件<span class="ff4">头</span>结构的定<span class="ff4">义</span>：</div><div class="t m0 x3 h9 y40 ff5 fs4 fc0 sc0 ls0 ws0"></div><div class="t m0 x3 h9 y41 ff5 fs4 fc0 sc0 ls0 ws0"></div><div class="t m0 x3 h9 y42 ff5 fs4 fc0 sc0 ls0 ws0"> !!"</div><div class="t m0 x3 h9 y43 ff5 fs4 fc0 sc0 ls0 ws0"> !#$%"</div><div class="t m0 x3 h9 y44 ff5 fs4 fc0 sc0 ls0 ws0"> !!&'"</div><div class="t m0 x3 h9 y45 ff5 fs4 fc0 sc0 ls0 ws0"> !!&("</div><div class="t m0 x3 h9 y46 ff5 fs4 fc0 sc0 ls0 ws0"> ! $"</div><div class="t m0 x3 h9 y47 ff5 fs4 fc0 sc0 ls0 ws0">)"</div><div class="t m0 x3 h8 y48 ff1 fs2 fc0 sc0 ls0 ws0">其中的<span class="ff3">bfT<span class="_ _3"></span>y<span class="_ _3"></span>pe<span class="ff1">值应该是<span class="ff4">“</span></span>BM”<span class="ff1">（</span>0x4d42<span class="ff1">），<span class="ff4">标志</span>该<span class="ff4">文</span>件是<span class="ff4">位</span>图<span class="ff4">文</span>件。<span class="_ _2"> </span></span>bfSize<span class="ff1">的值是<span class="ff4">位</span>图<span class="ff4">文</span></span></span></div><div class="t m0 x2 h8 y49 ff1 fs2 fc0 sc0 ls0 ws0">件的大<span class="ff4">小</span>。</div><div class="t m0 x3 h8 y4a ff4 fs2 fc0 sc1 ls0 ws0">位<span class="ff1">图信息</span>头</div><div class="t m0 x3 h8 y4b ff4 fs2 fc0 sc0 ls0 ws0">位<span class="ff1">图信息</span>头<span class="ff1">包含了单个像</span>素<span class="ff1">所用字</span>节<span class="ff1">数以及</span>描<span class="ff1">述</span>颜色<span class="ff1">的</span>格式<span class="ff1">，</span>此外<span class="ff1">还包括</span>位<span class="ff1">图的</span>宽<span class="ff1">度、高</span></div><div class="t m0 x2 h8 y4c ff1 fs2 fc0 sc0 ls0 ws0">度、目<span class="ff4">标</span>设<span class="ff4">备</span>的<span class="ff4">位</span>平面数、图像的<span class="ff4">压</span>缩<span class="ff4">格式</span>。</div><div class="t m0 x3 h8 y4d ff1 fs2 fc0 sc0 ls0 ws0">以下是<span class="ff4">位</span>图信息<span class="ff4">头</span>结构的定<span class="ff4">义</span>：</div><div class="t m0 x3 h9 y4e ff5 fs4 fc0 sc0 ls0 ws0">* $</div><div class="t m0 x3 h9 y4f ff5 fs4 fc0 sc0 ls0 ws0">!! !$#$%"</div><div class="t m0 x3 h9 y50 ff5 fs4 fc0 sc0 ls0 ws0">!! *+!$$"</div><div class="t m0 x3 h9 y51 ff5 fs4 fc0 sc0 ls0 ws0">! *+!$$"!</div><div class="t m0 x3 h9 y52 ff5 fs4 fc0 sc0 ls0 ws0">!! !$,-"</div><div class="t m0 x3 h9 y53 ff5 fs4 fc0 sc0 ls0 ws0">!! !$$./-</div><div class="t m0 x3 h9 y54 ff5 fs4 fc0 sc0 ls0 ws0">!! !$./$/-"</div><div class="t m0 x3 h9 y55 ff5 fs4 fc0 sc0 ls0 ws0">!! !$#$%"</div><div class="t m0 x3 h9 y56 ff5 fs4 fc0 sc0 ls0 ws0">!! *+!$0,"</div><div class="t m0 x3 h9 y57 ff5 fs4 fc0 sc0 ls0 ws0">!! *+!$1,"</div><div class="t m0 x3 h9 y58 ff5 fs4 fc0 sc0 ls0 ws0">!! !$.,2"</div><div class="t m0 x3 h9 y59 ff5 fs4 fc0 sc0 ls0 ws0">!! !$.,/-"</div></div></div><div class="pi" data-data='{"ctm":[1.839080,0.000000,0.000000,1.839080,0.000000,0.000000]}'></div></div>

<div id="pf5" class="pf w0 h0" data-page-no="5"><div class="pc pc5 w0 h0"><img class="bi x0 y0 w1 h1" alt="" src="https://static.pudn.com/prod/directory_preview_static/62781aead7996104df5759ff/bg5.jpg"><div class="c x0 y1 w2 h2"><div class="t m0 x3 h9 y5a ff5 fs4 fc0 sc0 ls0 ws0">)* "</div><div class="t m0 x2 h8 y5b ff1 fs2 fc0 sc0 ls0 ws0">下面是对结构体当中<span class="ff4">各</span>个成<span class="ff4">员</span>的说明：</div><div class="t m0 x2 ha y5c ff3 fs2 fc0 sc0 ls0 ws0">biSize </div><div class="t m0 x3 h8 y5d ff1 fs2 fc0 sc0 ls0 ws0">结构<span class="ff3">BITMAPINFOHEADER</span>的字<span class="ff4">节</span>数，<span class="ff4">即<span class="ff3">sizeof(BITMAPINFOHEADER)* </span></span></div><div class="t m0 x2 ha y5e ff3 fs2 fc0 sc0 ls0 ws0">biW<span class="_ _3"></span>idth </div><div class="t m0 x2 h8 y5f ff3 fs2 fc0 sc0 ls0 ws0"> <span class="ff1">以像<span class="ff4">素</span>为单<span class="ff4">位</span>的图像<span class="ff4">宽</span>度</span>* </div><div class="t m0 x2 ha y60 ff3 fs2 fc0 sc0 ls0 ws0">biHeight </div><div class="t m0 x3 h8 y61 ff1 fs2 fc0 sc0 ls0 ws0">以像<span class="ff4">素</span>为单<span class="ff4">位</span>的图像长度<span class="ff3">* </span></div><div class="t m0 x2 ha y62 ff3 fs2 fc0 sc0 ls0 ws0">biplanes </div><div class="t m0 x2 h8 y63 ff3 fs2 fc0 sc0 ls0 ws0"> <span class="ff1">目<span class="ff4">标</span>设<span class="ff4">备</span>的<span class="ff4">位</span>平面数</span> </div><div class="t m0 x2 ha y64 ff3 fs2 fc0 sc0 ls0 ws0">biBitCount </div><div class="t m0 x2 h8 y65 ff3 fs2 fc0 sc0 ls0 ws0"> <span class="ff1">每个像<span class="ff4">素</span>的<span class="ff4">位</span>数</span> </div><div class="t m0 x2 ha y66 ff3 fs2 fc0 sc0 ls0 ws0">biCompression </div><div class="t m0 x2 h8 y67 ff3 fs2 fc0 sc0 ls0 ws0"> <span class="ff1">图像的<span class="ff4">压</span>缩<span class="ff4">格式</span>（这个值<span class="ff4">几乎</span>总是为</span>0<span class="ff1">）</span> </div><div class="t m0 x2 ha y68 ff3 fs2 fc0 sc0 ls0 ws0">biSizeImage </div><div class="t m0 x2 h8 y69 ff3 fs2 fc0 sc0 ls0 ws0"> <span class="ff1">以字<span class="ff4">节</span>为单<span class="ff4">位</span>的图像数据的大<span class="ff4">小</span>（对</span>BI_RGB<span class="ff4">压<span class="ff1">缩方</span>式<span class="ff1">而</span>言<span class="ff1">）</span></span> </div><div class="t m0 x2 ha y6a ff3 fs2 fc0 sc0 ls0 ws0">biXPelsPermeter </div><div class="t m0 x2 h8 y6b ff3 fs2 fc0 sc0 ls0 ws0"> <span class="ff4">水<span class="ff1">平方</span>向<span class="ff1">上的每</span>米<span class="ff1">的像</span>素<span class="ff1">个数</span></span> </div><div class="t m0 x2 ha y6c ff3 fs2 fc0 sc0 ls0 ws0">biYPelsPerMeter </div><div class="t m0 x2 h8 y49 ff3 fs2 fc0 sc0 ls0 ws0"> <span class="ff4">垂直<span class="ff1">方</span>向<span class="ff1">上的每</span>米<span class="ff1">的像</span>素<span class="ff1">个数</span></span> </div><div class="t m0 x2 ha y6d ff3 fs2 fc0 sc0 ls0 ws0">biClrused </div><div class="t m0 x2 h8 y4a ff3 fs2 fc0 sc0 ls0 ws0"> <span class="ff1">调<span class="ff4">色板</span>中实<span class="ff4">际</span>使用的<span class="ff4">颜色</span>数</span></div><div class="t m0 x3 h8 y4b ff1 fs2 fc0 sc0 ls0 ws0">这个值通<span class="ff4">常</span>为<span class="ff3">0</span>，<span class="ff4">表</span>示使用<span class="ff3">biBitCount</span>确定的全部<span class="ff4">颜色</span>，<span class="ff4">例外</span>是使用的<span class="ff4">颜色</span>的数目<span class="ff4">小</span>于<span class="ff4">制</span>定</div><div class="t m0 x2 h8 y4c ff1 fs2 fc0 sc0 ls0 ws0">的<span class="ff4">颜色深</span>度的<span class="ff4">颜色</span>数目的最大值。</div><div class="t m0 x2 ha y6e ff3 fs2 fc0 sc0 ls0 ws0">biClrImportant </div><div class="t m0 x2 h8 y6f ff3 fs2 fc0 sc0 ls0 ws0"> <span class="ff1">现实<span class="ff4">位</span>图时<span class="ff4">必须</span>的<span class="ff4">颜色</span>数</span></div><div class="t m0 x3 h8 y70 ff1 fs2 fc0 sc0 ls0 ws0">这个值通<span class="ff4">常</span>为<span class="ff3">0</span>，<span class="ff4">表</span>示所有的<span class="ff4">颜色</span>都是<span class="ff4">必</span>需的</div><div class="t m0 x2 h8 y71 ff1 fs2 fc0 sc0 ls0 ws0">对于其中的<span class="ff3">biBitCount</span>，分别有以下<span class="ff4">意义</span>：<span class="ff3"> </span></div><div class="t m0 x3 h8 y72 ff3 fs2 fc0 sc0 ls0 ws0">0<span class="ff1">，用在</span>JPEG<span class="ff4">格式<span class="ff1">中</span></span> </div><div class="t m0 x3 h8 y73 ff3 fs2 fc0 sc0 ls0 ws0">1<span class="ff1">，单<span class="ff4">色</span>图，调<span class="ff4">色板</span>中含有两<span class="ff4">种颜色</span>，也就是我们通<span class="ff4">常</span>说的<span class="ff4">黑白</span>图片</span></div><div class="t m0 x3 h8 y74 ff3 fs2 fc0 sc0 ls0 ws0">4<span class="ff1">，</span>16<span class="ff4">色<span class="ff1">图</span></span> </div><div class="t m0 x3 h8 y75 ff3 fs2 fc0 sc0 ls0 ws0">8<span class="ff1">，</span>256<span class="ff4">色<span class="ff1">图，通</span>常<span class="ff1">说的灰度图</span></span></div><div class="t m0 x3 h8 y76 ff3 fs2 fc0 sc0 ls0 ws0">16<span class="ff1">，</span>64K<span class="ff1">图，一<span class="ff4">般没</span>有调<span class="ff4">色板</span>，图像数据中每两个字<span class="ff4">节表</span>示一个像<span class="ff4">素</span>，</span>5<span class="ff1">个或</span>6<span class="ff1">个<span class="ff4">位表</span>示一</span></div><div class="t m0 x2 h5 y77 ff1 fs2 fc0 sc0 ls0 ws0">个<span class="ff3">RGB</span>分量<span class="ff3"> </span></div><div class="t m0 x3 h8 y78 ff3 fs2 fc0 sc0 ls0 ws0">24<span class="ff1">，</span>16M<span class="ff4">真彩色<span class="ff1">图，一</span>般没<span class="ff1">有调</span>色板<span class="ff1">，图像数据中每</span></span>3<span class="ff1">个字<span class="ff4">节表</span>示一个像<span class="ff4">素</span>，每个字<span class="ff4">节表</span></span></div><div class="t m0 x2 h5 y79 ff1 fs2 fc0 sc0 ls0 ws0">示一个<span class="ff3">RGB</span>分量<span class="ff3"> </span></div></div></div><div class="pi" data-data='{"ctm":[1.839080,0.000000,0.000000,1.839080,0.000000,0.000000]}'></div></div>

BP神经网络算法完整学习资料 图象识别技术 入门必备

BP神经网络算法完整学习资料图象识别技术入门必备