【数字识别】基于知识库实现手写体数字识别附matlab代码

1 内容介绍

摘要：随着科学技术的发展，机器学习成为一大学科热门领域，是一门专门研究计算机怎样模拟或实现人类的学习行为的交叉学科。文章在matlab软件的基础上，利用BP神经网络算法完成手写体数字的识别。

关键词：机器学习；手写体数字识别；BP神经网络

机器学习是一门多领域交叉学科，专门研究计算机怎样模拟或实现人类的学习行为，以获取新的知识或技能，重新组织己有的知识结构，使之不断改善自身的性能。实现机器学习的方法多种多样，常见到的主要有神经网络算法、CNN卷积神经网络算法、RNN循环神经网络算法、EM算法、贝叶斯算法、聚类算法、回归算法、SVM等。本文将利用知识库算法来完成手写体数字的识别[1]。

​

2 仿真代码

<span style="color:#333333"><span style="background-color:rgba(0, 0, 0, 0.03)"><code>clc; clear all; close all;</code><code>load Data.mat;</code><code>[FileName,PathName,FilterIndex] = uigetfile({<span style="color:#dd1144">'*.jpg;*.tif;*.png;*.gif'</span>, ...</code><code>    <span style="color:#dd1144">'所有图像文件'</span>;...</code><code>    <span style="color:#dd1144">'*.*'</span>,<span style="color:#dd1144">'所有文件'</span> },<span style="color:#dd1144">'载入数字图像'</span>,...</code><code>    <span style="color:#dd1144">'.\\images\\手写数字\\t0.jpg'</span>);</code><code><span style="color:#ca7d37">if</span> isequal(FileName, <span style="color:#0e9ce5">0</span>) || isequal(PathName, <span style="color:#0e9ce5">0</span>)</code><code>    <span style="color:#ca7d37">return</span>;</code><code><span style="color:#ca7d37">end</span></code><code>fileName = fullfile(PathName, FileName);</code><code>I = imread(fileName);</code><code>flag = <span style="color:#0e9ce5">1</span>;</code><code>I1 = Normalize_Img(I);</code><code>bw1 = Bw_Img(I1);</code><code>bw2 = Thin_Img(bw1);</code><code>bw = bw2;</code><code>sz = size(bw);</code><code>[r, c] = find(bw==<span style="color:#0e9ce5">1</span>);</code><code>rect = [min(c) min(r) max(c)-min(c) max(r)-min(r)];</code><code>vs = rect(<span style="color:#0e9ce5">1</span>)+rect(<span style="color:#0e9ce5">3</span>)*[<span style="color:#0e9ce5">5</span>/<span style="color:#0e9ce5">12</span> <span style="color:#0e9ce5">1</span>/<span style="color:#0e9ce5">2</span> <span style="color:#0e9ce5">7</span>/<span style="color:#0e9ce5">12</span>];</code><code>hs = rect(<span style="color:#0e9ce5">2</span>)+rect(<span style="color:#0e9ce5">4</span>)*[<span style="color:#0e9ce5">1</span>/<span style="color:#0e9ce5">3</span> <span style="color:#0e9ce5">1</span>/<span style="color:#0e9ce5">2</span> <span style="color:#0e9ce5">2</span>/<span style="color:#0e9ce5">3</span>];</code><code>pt1 = [rect(<span style="color:#0e9ce5">1</span><span style="color:#dd1144">:</span><span style="color:#0e9ce5">2</span>); rect(<span style="color:#0e9ce5">1</span><span style="color:#dd1144">:</span><span style="color:#0e9ce5">2</span>)+rect(<span style="color:#0e9ce5">3</span><span style="color:#dd1144">:</span><span style="color:#0e9ce5">4</span>)];</code><code>pt2 = [rect(<span style="color:#0e9ce5">1</span>)+rect(<span style="color:#0e9ce5">3</span>) rect(<span style="color:#0e9ce5">2</span>); rect(<span style="color:#0e9ce5">1</span>) rect(<span style="color:#0e9ce5">2</span>)+rect(<span style="color:#0e9ce5">4</span>)];</code><code>k1 = (pt1(<span style="color:#0e9ce5">1</span>,<span style="color:#0e9ce5">2</span>)-pt1(<span style="color:#0e9ce5">2</span>,<span style="color:#0e9ce5">2</span>)) / (pt1(<span style="color:#0e9ce5">1</span>,<span style="color:#0e9ce5">1</span>)-pt1(<span style="color:#0e9ce5">2</span>,<span style="color:#0e9ce5">1</span>));</code><code>x1 = <span style="color:#0e9ce5">1</span><span style="color:#dd1144">:sz</span>(<span style="color:#0e9ce5">2</span>);</code><code>y1 = k1*(x1-pt1(<span style="color:#0e9ce5">1</span>,<span style="color:#0e9ce5">1</span>)) + pt1(<span style="color:#0e9ce5">1</span>,<span style="color:#0e9ce5">2</span>);</code><code>k2 = (pt2(<span style="color:#0e9ce5">1</span>,<span style="color:#0e9ce5">2</span>)-pt2(<span style="color:#0e9ce5">2</span>,<span style="color:#0e9ce5">2</span>)) / (pt2(<span style="color:#0e9ce5">1</span>,<span style="color:#0e9ce5">1</span>)-pt2(<span style="color:#0e9ce5">2</span>,<span style="color:#0e9ce5">1</span>));</code><code>x2 = <span style="color:#0e9ce5">1</span><span style="color:#dd1144">:sz</span>(<span style="color:#0e9ce5">2</span>);</code><code>y2 = k2*(x2-pt2(<span style="color:#0e9ce5">1</span>,<span style="color:#0e9ce5">1</span>)) + pt2(<span style="color:#0e9ce5">1</span>,<span style="color:#0e9ce5">2</span>);</code><code><span style="color:#ca7d37">if</span> flag</code><code>    figure(<span style="color:#dd1144">'Name'</span>, <span style="color:#dd1144">'数字识别'</span>, <span style="color:#dd1144">'NumberTitle'</span>, <span style="color:#dd1144">'Off'</span>, <span style="color:#dd1144">'Units'</span>, <span style="color:#dd1144">'Normalized'</span>, <span style="color:#dd1144">'Position'</span>, [<span style="color:#0e9ce5">0</span>.<span style="color:#0e9ce5">2</span> <span style="color:#0e9ce5">0</span>.<span style="color:#0e9ce5">45</span> <span style="color:#0e9ce5">0</span>.<span style="color:#0e9ce5">5</span> <span style="color:#0e9ce5">0</span>.<span style="color:#0e9ce5">3</span>]);</code><code>    subplot(<span style="color:#0e9ce5">2</span>, <span style="color:#0e9ce5">2</span>, <span style="color:#0e9ce5">1</span>); imshow(I, []); title(<span style="color:#dd1144">'原图像'</span>, <span style="color:#dd1144">'FontWeight'</span>, <span style="color:#dd1144">'Bold'</span>);</code><code>    subplot(<span style="color:#0e9ce5">2</span>, <span style="color:#0e9ce5">2</span>, <span style="color:#0e9ce5">2</span>); imshow(I1, []); title(<span style="color:#dd1144">'归一化图像'</span>, <span style="color:#dd1144">'FontWeight'</span>, <span style="color:#dd1144">'Bold'</span>);</code><code>    hold on;</code><code>    h = rectangle(<span style="color:#dd1144">'Position'</span>, [rect(<span style="color:#0e9ce5">1</span><span style="color:#dd1144">:</span><span style="color:#0e9ce5">2</span>)-<span style="color:#0e9ce5">1</span> rect(<span style="color:#0e9ce5">3</span><span style="color:#dd1144">:</span><span style="color:#0e9ce5">4</span>)+<span style="color:#0e9ce5">2</span>], <span style="color:#dd1144">'EdgeColor'</span>, <span style="color:#dd1144">'r'</span>, <span style="color:#dd1144">'LineWidth'</span>, <span style="color:#0e9ce5">2</span>);</code><code>    xlabel(<span style="color:#dd1144">'数字区域标记'</span>);</code><code>    subplot(<span style="color:#0e9ce5">2</span>, <span style="color:#0e9ce5">2</span>, <span style="color:#0e9ce5">3</span>); imshow(bw1, []); title(<span style="color:#dd1144">'二值化图像'</span>, <span style="color:#dd1144">'FontWeight'</span>, <span style="color:#dd1144">'Bold'</span>);</code><code>    subplot(<span style="color:#0e9ce5">2</span>, <span style="color:#0e9ce5">2</span>, <span style="color:#0e9ce5">4</span>); imshow(bw, [], <span style="color:#dd1144">'Border'</span>, <span style="color:#dd1144">'Loose'</span>); title(<span style="color:#dd1144">'细化图像'</span>, <span style="color:#dd1144">'FontWeight'</span>, <span style="color:#dd1144">'Bold'</span>);</code><code>    hold on;</code><code>    h = [];</code><code>    <span style="color:#ca7d37">for</span> i = <span style="color:#0e9ce5">1</span> : length(hs)</code><code>        h = [h plot([<span style="color:#0e9ce5">1</span> sz(<span style="color:#0e9ce5">2</span>)], [hs(i) hs(i)], <span style="color:#dd1144">'r-'</span>)];</code><code>    <span style="color:#ca7d37">end</span></code><code>    <span style="color:#ca7d37">for</span> i = <span style="color:#0e9ce5">1</span> : length(vs)</code><code>        h = [h plot([vs(i) vs(i)], [<span style="color:#0e9ce5">1</span> sz(<span style="color:#0e9ce5">1</span>)], <span style="color:#dd1144">'g-'</span>)];</code><code>    <span style="color:#ca7d37">end</span></code><code>    h = [h plot(x1, y1, <span style="color:#dd1144">'y-'</span>)];</code><code>    h = [h plot(x2, y2, <span style="color:#dd1144">'m-'</span>)];</code><code>    legend([h(<span style="color:#0e9ce5">1</span>) h(<span style="color:#0e9ce5">4</span>) h(<span style="color:#0e9ce5">7</span>) h(<span style="color:#0e9ce5">8</span>)], {<span style="color:#dd1144">'水平线'</span>, <span style="color:#dd1144">'竖直线'</span>, <span style="color:#dd1144">'左对角线'</span>, <span style="color:#dd1144">'右对角线'</span>}, <span style="color:#dd1144">'Location'</span>, <span style="color:#dd1144">'BestOutside'</span>);</code><code>    hold off;</code><code><span style="color:#ca7d37">end</span></code><code>v{<span style="color:#0e9ce5">1</span>} = [<span style="color:#0e9ce5">1</span><span style="color:#dd1144">:sz</span>(<span style="color:#0e9ce5">2</span>); repmat(hs(<span style="color:#0e9ce5">1</span>), <span style="color:#0e9ce5">1</span>, sz(<span style="color:#0e9ce5">2</span>))]<span style="color:#dd1144">';</span></code><code>v{2} = [1:sz(2); repmat(hs(2), 1, sz(2))]';</code><code>v{<span style="color:#0e9ce5">3</span>} = [<span style="color:#0e9ce5">1</span><span style="color:#dd1144">:sz</span>(<span style="color:#0e9ce5">2</span>); repmat(hs(<span style="color:#0e9ce5">3</span>), <span style="color:#0e9ce5">1</span>, sz(<span style="color:#0e9ce5">2</span>))]<span style="color:#dd1144">';</span></code><code>v{4} = [repmat(vs(1), 1, sz(1)); 1:sz(1)]';</code><code>v{<span style="color:#0e9ce5">5</span>} = [repmat(vs(<span style="color:#0e9ce5">2</span>), <span style="color:#0e9ce5">1</span>, sz(<span style="color:#0e9ce5">1</span>)); <span style="color:#0e9ce5">1</span><span style="color:#dd1144">:sz</span>(<span style="color:#0e9ce5">1</span>)]<span style="color:#dd1144">';</span></code><code>v{6} = [repmat(vs(3), 1, sz(1)); 1:sz(1)]';</code><code>v{<span style="color:#0e9ce5">7</span>} = [x1; y1]<span style="color:#dd1144">';</span></code><code>v{8} = [x2; y2]';</code><code><span style="color:#ca7d37">for</span> i = <span style="color:#0e9ce5">1</span> : <span style="color:#0e9ce5">8</span></code><code>    num(i) = GetImgLinePts(bw, round(v{i})-<span style="color:#0e9ce5">1</span>);</code><code><span style="color:#ca7d37">end</span></code><code>num(<span style="color:#0e9ce5">9</span>) = sum(sum(endpoints(bw)));</code><code>result = MaskRecon(Datas, num);</code><code>msgbox(sprintf(<span style="color:#dd1144">'识别结果:%d'</span>, result), <span style="color:#dd1144">'提示信息'</span>, <span style="color:#dd1144">'modal'</span>);</code><code>​</code></span></span>

3 运行结果

4 参考文献

[1]刘思慧, 江维. 基于MATLAB手写体数字识别程序设计[J]. 电子世界, 2019(3):2.

[2]丁禹鑫, 丁会, 张红娟,等. 基于matlab的手写体数字识别系统研究[J]. 无线互联科技, 2017(18):2.

[3]张鱼米. 自制人工智能:用神经网络识别手写数字[J].  2018.

博主简介：擅长智能优化算法、神经网络预测、信号处理、元胞自动机、图像处理、路径规划、无人机等多种领域的Matlab仿真，相关matlab代码问题可私信交流。

部分理论引用网络文献，若有侵权联系博主删除。