首先推荐几篇有关验证码识别的文章,觉得不错
php实现验证码的识别(初级篇)
关于bp神经网格识别验证码
一、思路
碰见一个验证码，如果我们想要识别它，我们需要的是做什么呢？
我们先观察几个验证码............
我们用人眼去观察，会很显然的认出验证码所包含的字符，那么人眼的“识别机理”是什么呢？
大概是验证码图片字符的背景的颜色区别吧，试想，如果字符和背景没有颜色区别，我们能够判断验证码吗，很显然不能。
所以，我们就可以从人出发。
先从图片的颜色着手，即图片的RGB信息。
RGB色彩模式是工业界的一种颜色标准，是通过对红(R)、绿(G)、蓝(B)三个颜色通道的变化以及它们相互之间的叠加来得到各式各样的颜色的，RGB即是代表红、绿、蓝三个通道的颜色，这个标准几乎包括了人类视力所能感知的所有颜色，是目前运用最广的颜色系统之一。
定义函数取得RGB信息
<span> 1</span> <span>//</span><span>代码本来是一个类，现在拆开来写的，有可能有不严谨的地方，大家可以看得懂就好了</span>
<span> 2</span>
<span> 3</span> <span>/*</span>
<span> 4</span> <span> *取得图片路径和图片尺寸
</span><span> 5</span> <span>*/</span>
<span> 6</span> <span>$this</span>->ImagePath = <span>$Image</span><span>;
</span><span> 7</span> <span>$this</span>->ImageSize = <span>getimagesize</span>(<span>$Image</span><span>);
</span><span> 8</span>
<span> 9</span> <span>/*</span>
<span>10</span> <span> *获取图像标识符，保存到ImageInfo，只能处理bmp,png,jpg图片
</span><span>11</span> <span> *ImageCreateFromBmp是我自己定义的函数，最后会给出
</span><span>12</span> <span>*/</span>
<span>13</span> <span>function</span><span> getInfo(){
</span><span>14</span> <span>$filetype</span> = <span>substr</span>(<span>$this</span>->ImagePath,-3<span>);
</span><span>15</span> <span>if</span>(<span>$filetype</span> == 'bmp'<span>){
</span><span>16</span> <span>$this</span>->ImageInfo = <span>$this</span>->ImageCreateFromBmp(<span>$this</span>-><span>ImagePath);
</span><span>17</span> }<span>elseif</span>(<span>$filetype</span> == 'jpg'<span>){
</span><span>18</span> <span>$this</span>->ImageInfo = imagecreatefromjpeg(<span>$this</span>-><span>ImagePath);
</span><span>19</span> }<span>elseif</span>(<span>$filetype</span> == 'png'<span>){
</span><span>20</span> <span>$this</span>->ImageInfo = imagecreatefrompng(<span>$this</span>-><span>ImagePath);
</span><span>21</span> <span> }
</span><span>22</span> <span>}
</span><span>23</span>
<span>24</span> <span>/*</span><span>获取图片RGB信息</span><span>*/</span>
<span>25</span> <span>function</span><span> getRgb(){
</span><span>26</span> <span>$rgbArray</span> = <span>array</span><span>();
</span><span>27</span> <span>$res</span> = <span>$this</span>-><span>ImageInfo;
</span><span>28</span> <span>$size</span> = <span>$this</span>-><span>ImageSize;
</span><span>29</span> <span>$wid</span> = <span>$size</span>['0'<span>];
</span><span>30</span> <span>$hid</span> = <span>$size</span>['1'<span>];
</span><span>31</span> <span>for</span>(<span>$i</span>=0; <span>$i</span> < <span>$hid</span>; ++<span>$i</span><span>){
</span><span>32</span> <span>for</span>(<span>$j</span>=0; <span>$j</span> < <span>$wid</span>; ++<span>$j</span><span>){
</span><span>33</span> <span>$rgb</span> = imagecolorat(<span>$res</span>,<span>$j</span>,<span>$i</span><span>);
</span><span>34</span> <span>$rgbArray</span>[<span>$i</span>][<span>$j</span>] = imagecolorsforindex(<span>$res</span>, <span>$rgb</span><span>);
</span><span>35</span> <span> }
</span><span>36</span> <span> }
</span><span>37</span> <span>return</span> <span>$rgbArray</span><span>;
</span><span>38</span> }
二、二值化
因为人眼可以分别出验证码，所以验证码的RGB信息就会有一定的特点，这时候需要我们观察一下，直接打印RGB数组是不好观察的…………，好多数啊
在php实现验证码的识别(初级篇)中，作者的判断依据是
无论验证数字颜色如何变化，该数字的 RGB 值总有一个值小于 125
我们先获取他的灰度，再判断
<span> 1</span> <span>/*</span>
<span> 2</span> <span> *获取灰度信息
</span><span> 3</span> <span>*/</span>
<span> 4</span> <span>function</span><span> getGray(){
</span><span> 5</span> <span>$grayArray</span> = <span>array</span><span>();
</span><span> 6</span> <span>$size</span> = <span>$this</span>-><span>ImageSize;
</span><span> 7</span> <span>$rgbarray</span> = <span>$this</span>-><span>getRgb();
</span><span> 8</span> <span>$wid</span> = <span>$size</span>['0'<span>];
</span><span> 9</span> <span>$hid</span> = <span>$size</span>['1'<span>];
</span><span>10</span> <span>for</span>(<span>$i</span>=0; <span>$i</span> < <span>$hid</span>; ++<span>$i</span><span>){
</span><span>11</span> <span>for</span>(<span>$j</span>=0; <span>$j</span> < <span>$wid</span>; ++<span>$j</span><span>){
</span><span>12</span> <span>$grayArray</span>[<span>$i</span>][<span>$j</span>] = (299*<span>$rgbarray</span>[<span>$i</span>][<span>$j</span>]['red']+587*<span>$rgbarray</span>[<span>$i</span>][<span>$j</span>]['green']+144*<span>$rgbarray</span>[<span>$i</span>][<span>$j</span>]['blue'])/1000<span>;
</span><span>13</span> <span> }
</span><span>14</span> <span> }
</span><span>15</span> <span>return</span> <span>$grayArray</span><span>;
</span><span>16</span> }
然后我们根据灰度信息，打印图片，注意不是打印灰度信息
<span> 1</span> <span>/*</span><span>根据灰度信息打印图片</span><span>*/</span>
<span> 2</span> <span>function</span><span> printByGray(){
</span><span> 3</span> <span>$size</span> = <span>$this</span>-><span>ImageSize;
</span><span> 4</span> <span>$grayArray</span> = <span>$this</span>-><span>getGray();
</span><span> 5</span> <span>$wid</span> = <span>$size</span>['0'<span>];
</span><span> 6</span> <span>$hid</span> = <span>$size</span>['1'<span>];
</span><span> 7</span> <span>for</span>(<span>$k</span>=0;<span>$k</span><25;<span>$k</span>++<span>){
</span><span> 8</span> <span>echo</span> <span>$k</span>."\n"<span>;
</span><span> 9</span> <span>for</span>(<span>$i</span>=0; <span>$i</span> < <span>$hid</span>; ++<span>$i</span><span>){
</span><span>10</span> <span>for</span>(<span>$j</span>=0; <span>$j</span> < <span>$wid</span>; ++<span>$j</span><span>){
</span><span>11</span> <span>if</span>(<span>$grayArray</span>[<span>$i</span>][<span>$j</span>] < <span>$k</span>*10<span>){
</span><span>12</span> <span>echo</span> '■'<span>;
</span><span>13</span> }<span>else</span><span>{
</span><span>14</span> <span>echo</span> '□'<span>;
</span><span>15</span> <span> }
</span><span>16</span> <span> }
</span><span>17</span> <span>echo</span> "|\n"<span>;
</span><span>18</span> <span> }
</span><span>19</span> <span>echo</span> "---------------------------------------------------------------------------------------------------------------\n"<span>;
</span><span>20</span> <span> }
</span><span>21</span>
<span>22</span> }
注意到，从$grayArray[$i][$j] < 80就会有显然的输出，我们观察选择了一个恰当的阈值，得到一个0101的数组，我们已经将我们的图片转化为了字符（1）和背景（0），即二值化。
<span> 1</span> <span>/*</span>
<span> 2</span> <span> *根据自定义的规则，获取二值化二维数组
</span><span> 3</span> <span> *@return 图片高*宽的二值数组（0,1）
</span><span> 4</span> <span>*/</span>
<span> 5</span> <span>function</span><span> getErzhi(){
</span><span> 6</span> <span>$erzhiArray</span> = <span>array</span><span>();
</span><span> 7</span> <span>$size</span> = <span>$this</span>-><span>ImageSize;
</span><span> 8</span> <span>$grayArray</span> = <span>$this</span>-><span>getGray();
</span><span> 9</span> <span>$wid</span> = <span>$size</span>['0'<span>];
</span><span>10</span> <span>$hid</span> = <span>$size</span>['1'<span>];
</span><span>11</span> <span>for</span>(<span>$i</span>=0; <span>$i</span> < <span>$hid</span>; ++<span>$i</span><span>){
</span><span>12</span> <span>for</span>(<span>$j</span>=0; <span>$j</span> <<span>$wid</span>; ++<span>$j</span><span>){
</span><span>13</span> <span>if</span>( <span>$grayArray</span>[<span>$i</span>][<span>$j</span>] < 90<span> ){
</span><span>14</span> <span>$erzhiArray</span>[<span>$i</span>][<span>$j</span>]=1<span>;
</span><span>15</span> }<span>else</span><span>{
</span><span>16</span> <span>$erzhiArray</span>[<span>$i</span>][<span>$j</span>]=0<span>;
</span><span>17</span> <span> }
</span><span>18</span> <span> }
</span><span>19</span> <span> }
</span><span>20</span> <span>return</span> <span>$erzhiArray</span><span>;
</span><span>21</span> }
三、去除噪点
但是我们发现有一些小点影响了我们的判断
我们可以注意到这些事干扰噪点，但是如果我们是机器的话，我们如何判断这些点是否是字符呢？
所以接下来，我们需要将这些字符去除。
我们判断，如果一个黑点的上下左右的八个点全部是白，我们就认为它是噪点，并予以清除，赋为白
<span> 1</span> <span>/*</span>
<span> 2</span> <span> *二值化图片降噪
</span><span> 3</span> <span> *@param $erzhiArray二值化数组
</span><span> 4</span> <span>*/</span>
<span> 5</span> <span>function</span> reduceZao(<span>$erzhiArray</span><span>){
</span><span> 6</span> <span>$data</span> = <span>$erzhiArray</span><span>;
</span><span> 7</span> <span>$gao</span> = <span>count</span>(<span>$erzhiArray</span><span>);
</span><span> 8</span> <span>$chang</span> = <span>count</span>(<span>$erzhiArray</span>['0'<span>]);
</span><span> 9</span>
<span>10</span> <span>$jiangzaoErzhiArray</span> = <span>array</span><span>();
</span><span>11</span>
<span>12</span> <span>for</span>(<span>$i</span>=0;<span>$i</span><<span>$gao</span>;<span>$i</span>++<span>){
</span><span>13</span> <span>for</span>(<span>$j</span>=0;<span>$j</span><<span>$chang</span>;<span>$j</span>++<span>){
</span><span>14</span> <span>$num</span> = 0<span>;
</span><span>15</span> <span>if</span>(<span>$data</span>[<span>$i</span>][<span>$j</span>] == 1<span>)
</span><span>16</span> <span> {
</span><span>17</span> <span>//</span><span> 上 </span>
<span>18</span> <span>if</span>(<span>isset</span>(<span>$data</span>[<span>$i</span>-1][<span>$j</span><span>])){
</span><span>19</span> <span>$num</span> = <span>$num</span> + <span>$data</span>[<span>$i</span>-1][<span>$j</span><span>];
</span><span>20</span> <span> }
</span><span>21</span> <span>//</span><span> 下 </span>
<span>22</span> <span>if</span>(<span>isset</span>(<span>$data</span>[<span>$i</span>+1][<span>$j</span><span>])){
</span><span>23</span> <span>$num</span> = <span>$num</span> + <span>$data</span>[<span>$i</span>+1][<span>$j</span><span>];
</span><span>24</span> <span> }
</span><span>25</span> <span>//</span><span> 左 </span>
<span>26</span> <span>if</span>(<span>isset</span>(<span>$data</span>[<span>$i</span>][<span>$j</span>-1<span>])){
</span><span>27</span> <span>$num</span> = <span>$num</span> + <span>$data</span>[<span>$i</span>][<span>$j</span>-1<span>];
</span><span>28</span> <span> }
</span><span>29</span> <span>//</span><span> 右 </span>
<span>30</span> <span>if</span>(<span>isset</span>(<span>$data</span>[<span>$i</span>][<span>$j</span>+1<span>])){
</span><span>31</span> <span>$num</span> = <span>$num</span> + <span>$data</span>[<span>$i</span>][<span>$j</span>+1<span>];
</span><span>32</span> <span> }
</span><span>33</span> <span>//</span><span> 上左 </span>
<span>34</span> <span>if</span>(<span>isset</span>(<span>$data</span>[<span>$i</span>-1][<span>$j</span>-1<span>])){
</span><span>35</span> <span>$num</span> = <span>$num</span> + <span>$data</span>[<span>$i</span>-1][<span>$j</span>-1<span>];
</span><span>36</span> <span> }
</span><span>37</span> <span>//</span><span> 上右 </span>
<span>38</span> <span>if</span>(<span>isset</span>(<span>$data</span>[<span>$i</span>-1][<span>$j</span>+1<span>])){
</span><span>39</span> <span>$num</span> = <span>$num</span> + <span>$data</span>[<span>$i</span>-1][<span>$j</span>+1<span>];
</span><span>40</span> <span> }
</span><span>41</span> <span>//</span><span> 下左 </span>
<span>42</span> <span>if</span>(<span>isset</span>(<span>$data</span>[<span>$i</span>+1][<span>$j</span>-1<span>])){
</span><span>43</span> <span>$num</span> = <span>$num</span> + <span>$data</span>[<span>$i</span>+1][<span>$j</span>-1<span>];
</span><span>44</span> <span> }
</span><span>45</span> <span>//</span><span> 下右 </span>
<span>46</span> <span>if</span>(<span>isset</span>(<span>$data</span>[<span>$i</span>+1][<span>$j</span>+1<span>])){
</span><span>47</span> <span>$num</span> = <span>$num</span> + <span>$data</span>[<span>$i</span>+1][<span>$j</span>+1<span>];
</span><span>48</span> <span> }
</span><span>49</span> <span> }
</span><span>50</span>
<span>51</span> <span>if</span>(<span>$num</span> < 1<span>){
</span><span>52</span> <span>$jiangzaoErzhiArray</span>[<span>$i</span>][<span>$j</span>] = 0<span>;
</span><span>53</span> }<span>else</span><span>{
</span><span>54</span> <span>$jiangzaoErzhiArray</span>[<span>$i</span>][<span>$j</span>] = 1<span>;
</span><span>55</span> <span> }
</span><span>56</span> <span> }
</span><span>57</span> <span> }
</span><span>58</span> <span>return</span> <span>$jiangzaoErzhiArray</span><span>;
</span><span>59</span>
<span>60</span> }
我们发现噪点消失了。
四、分割
这个时候，我们就需要对单一数字字母进行操作了，我们先将数字提取出来。
有些验证码字符相连，特别难!!!
我们分别从左到右，从右到左，从上到下，从下到上，进行扫描，去除白点，找到边框。
<span> 1</span> <span>/*</span>
<span> 2</span> <span> *归一化处理,针对一个个的数字,即去除字符周围的白点
</span><span> 3</span> <span> *@param $singleArray 二值化数组
</span><span> 4</span> <span>*/</span>
<span> 5</span> <span>function</span> getJinsuo(<span>$singleArray</span><span>){
</span><span> 6</span> <span>$dianCount</span> = 0<span>;
</span><span> 7</span> <span>$rearr</span> = <span>array</span><span>();
</span><span> 8</span>
<span> 9</span> <span>$gao</span> = <span>count</span>(<span>$singleArray</span><span>);
</span><span>10</span> <span>$kuan</span> = <span>count</span>(<span>$singleArray</span>['0'<span>]);
</span><span>11</span>
<span>12</span> <span>$dianCount</span> = 0<span>;
</span><span>13</span> <span>$shangKuang</span> = 0<span>;
</span><span>14</span> <span>$xiaKuang</span> = 0<span>;
</span><span>15</span> <span>$zuoKuang</span> = 0<span>;
</span><span>16</span> <span>$youKuang</span> = 0<span>;
</span><span>17</span> <span>//</span><span>从上到下扫描</span>
<span>18</span> <span>for</span>(<span>$i</span>=0; <span>$i</span> < <span>$gao</span>; ++<span>$i</span><span>){
</span><span>19</span> <span>for</span>(<span>$j</span>=0; <span>$j</span> < <span>$kuan</span>; ++<span>$j</span><span>){
</span><span>20</span> <span>if</span>( <span>$singleArray</span>[<span>$i</span>][<span>$j</span>] == 1<span>){
</span><span>21</span> <span>$dianCount</span>++<span>;
</span><span>22</span> <span> }
</span><span>23</span> <span> }
</span><span>24</span> <span>if</span>(<span>$dianCount</span>>1<span>){
</span><span>25</span> <span>$shangKuang</span> = <span>$i</span><span>;
</span><span>26</span> <span>$dianCount</span> = 0<span>;
</span><span>27</span> <span>break</span><span>;
</span><span>28</span> <span> }
</span><span>29</span> <span> }
</span><span>30</span> <span>//</span><span>从下到上扫描</span>
<span>31</span> <span>for</span>(<span>$i</span>=<span>$gao</span>-1; <span>$i</span> > -1; <span>$i</span>--<span>){
</span><span>32</span> <span>for</span>(<span>$j</span>=0; <span>$j</span> < <span>$kuan</span>; ++<span>$j</span><span>){
</span><span>33</span> <span>if</span>( <span>$singleArray</span>[<span>$i</span>][<span>$j</span>] == 1<span>){
</span><span>34</span> <span>$dianCount</span>++<span>;
</span><span>35</span> <span> }
</span><span>36</span> <span> }
</span><span>37</span> <span>if</span>(<span>$dianCount</span>>1<span>){
</span><span>38</span> <span>$xiaKuang</span> = <span>$i</span><span>;
</span><span>39</span> <span>$dianCount</span> = 0<span>;
</span><span>40</span> <span>break</span><span>;
</span><span>41</span> <span> }
</span><span>42</span> <span> }
</span><span>43</span> <span>//</span><span>从左到右扫描</span>
<span>44</span> <span>for</span>(<span>$i</span>=0; <span>$i</span> < <span>$kuan</span>; ++<span>$i</span><span>){
</span><span>45</span> <span>for</span>(<span>$j</span>=0; <span>$j</span> < <span>$gao</span>; ++<span>$j</span><span>){
</span><span>46</span> <span>if</span>( <span>$singleArray</span>[<span>$j</span>][<span>$i</span>] == 1<span>){
</span><span>47</span> <span>$dianCount</span>++<span>;
</span><span>48</span> <span> }
</span><span>49</span> <span> }
</span><span>50</span> <span>if</span>(<span>$dianCount</span>>1<span>){
</span><span>51</span> <span>$zuoKuang</span> = <span>$i</span><span>;
</span><span>52</span> <span>$dianCount</span> = 0<span>;
</span><span>53</span> <span>break</span><span>;
</span><span>54</span> <span> }
</span><span>55</span> <span> }
</span><span>56</span> <span>//</span><span>从右到左扫描</span>
<span>57</span> <span>for</span>(<span>$i</span>=<span>$kuan</span>-1; <span>$i</span> > -1; --<span>$i</span><span>){
</span><span>58</span> <span>for</span>(<span>$j</span>=0; <span>$j</span> < <span>$gao</span>; ++<span>$j</span><span>){
</span><span>59</span> <span>if</span>( <span>$singleArray</span>[<span>$j</span>][<span>$i</span>] == 1<span>){
</span><span>60</span> <span>$dianCount</span>++<span>;
</span><span>61</span> <span> }
</span><span>62</span> <span> }
</span><span>63</span> <span>if</span>(<span>$dianCount</span>>1<span>){
</span><span>64</span> <span>$youKuang</span> = <span>$i</span><span>;
</span><span>65</span> <span>$dianCount</span> = 0<span>;
</span><span>66</span> <span>break</span><span>;
</span><span>67</span> <span> }
</span><span>68</span> <span> }
</span><span>69</span> <span>for</span>(<span>$i</span>=0;<span>$i</span><<span>$xiaKuang</span>-<span>$shangKuang</span>+1;<span>$i</span>++<span>){
</span><span>70</span> <span>for</span>(<span>$j</span>=0;<span>$j</span><<span>$youKuang</span>-<span>$zuoKuang</span>+1;<span>$j</span>++<span>){
</span><span>71</span> <span>$rearr</span>[<span>$i</span>][<span>$j</span>] = <span>$singleArray</span>[<span>$shangKuang</span>+<span>$i</span>][<span>$zuoKuang</span>+<span>$j</span><span>];
</span><span>72</span> <span> }
</span><span>73</span> <span> }
</span><span>74</span> <span>return</span> <span>$rearr</span><span>;
</span><span>75</span> }
然后从左到右扫描，找到字符的分割
返回三维数组，每一维就是一个字符。
<span> 1</span> <span>/*</span>
<span> 2</span> <span> *切割成三维数组，每个小数字在一个数组里面
</span><span> 3</span> <span> *只适用四个数字一起的数组
</span><span> 4</span> <span> *@param 经过归一化处理的二值化数组
</span><span> 5</span> <span>*/</span>
<span> 6</span> <span>function</span> cutSmall(<span>$erzhiArray</span><span>){
</span><span> 7</span> <span>$doubleArray</span> = <span>array</span><span>();
</span><span> 8</span> <span>$jieZouyou</span> = <span>array</span><span>();
</span><span> 9</span>
<span>10</span> <span>$gao</span> = <span>count</span>(<span>$erzhiArray</span><span>);
</span><span>11</span> <span>$kuan</span> = <span>count</span>(<span>$erzhiArray</span>['0'<span>]);
</span><span>12</span>
<span>13</span> <span>$jie</span> = 0<span>;
</span><span>14</span> <span>$s</span> = 0<span>;
</span><span>15</span> <span>$jieZouyou</span>[<span>$s</span>] = 0<span>;
</span><span>16</span> <span>$s</span>++<span>;
</span><span>17</span> <span>//</span><span>从左到右扫描</span>
<span>18</span>
<span>19</span> <span>for</span>(<span>$i</span>=0; <span>$i</span> < <span>$kuan</span><span>;){
</span><span>20</span> <span>for</span>(<span>$j</span>=0; <span>$j</span> < <span>$gao</span>; ++<span>$j</span><span>){
</span><span>21</span> <span>$jie</span> = <span>$jie</span> + <span>$erzhiArray</span>[<span>$j</span>][<span>$i</span><span>];
</span><span>22</span> <span> }
</span><span>23</span> <span>//</span><span>如果有一列全部是白，设置$jieZouyou,并且跳过中间空白部分</span>
<span>24</span> <span>if</span>(<span>$jie</span> == 0<span>){
</span><span>25</span> <span>$jieZouyou</span>[<span>$s</span>] = <span>$i</span>+1<span>;
</span><span>26</span> <span>do</span><span>{
</span><span>27</span> <span>$n</span> = ++<span>$i</span><span>;
</span><span>28</span> <span>$qian</span> = 0<span>;
</span><span>29</span> <span>$hou</span> = 0<span>;
</span><span>30</span> <span>for</span>(<span>$m</span>=0; <span>$m</span> < <span>$gao</span>; ++<span>$m</span><span>){
</span><span>31</span> <span>$qian</span> = <span>$qian</span> + <span>$erzhiArray</span>[<span>$m</span>][<span>$n</span><span>];
</span><span>32</span> <span>$hou</span> = <span>$hou</span> + <span>$erzhiArray</span>[<span>$m</span>][<span>$n</span>+1<span>];
</span><span>33</span> <span> }
</span><span>34</span> <span>$jieZouyou</span>[<span>$s</span>+1] = <span>$n</span>+1<span>;
</span><span>35</span> <span> }
</span><span>36</span> <span>//</span><span>当有两列同时全部为白，说明有间隙，循环，知道间隙没有了</span>
<span>37</span> <span>while</span>(<span>$qian</span> == 0 && <span>$hou</span> == 0<span>);
</span><span>38</span> <span>$s</span>+=2<span>;
</span><span>39</span> <span>$i</span>++<span>;
</span><span>40</span> }<span>else</span><span>{
</span><span>41</span> <span>$i</span>++<span>;
</span><span>42</span> <span> }
</span><span>43</span>
<span>44</span> <span>$jie</span> = 0<span>;
</span><span>45</span> <span> }
</span><span>46</span> <span>$jieZouyou</span>[] = <span>$kuan</span><span>;
</span><span>47</span> <span>//</span><span>极端节点数量，(应该是字符个数)*2</span>
<span>48</span> <span>$jieZouyouCount</span> = <span>count</span>(<span>$jieZouyou</span><span>);
</span><span>49</span>
<span>50</span> <span>for</span>(<span>$k</span>=0;<span>$k</span><<span>$jieZouyouCount</span>/2;<span>$k</span>++<span>){
</span><span>51</span> <span>for</span>(<span>$i</span>=0; <span>$i</span> < <span>$gao</span>; <span>$i</span>++<span>){
</span><span>52</span> <span>for</span>(<span>$j</span>=0; <span>$j</span> < <span>$jieZouyou</span>[<span>$k</span>*2+1]-<span>$jieZouyou</span>[<span>$k</span>*2]-1; ++<span>$j</span><span>){
</span><span>53</span> <span>$doubleArray</span>[<span>$k</span>][<span>$i</span>][<span>$j</span>] = <span>$erzhiArray</span>[<span>$i</span>][<span>$j</span>+<span>$jieZouyou</span>[<span>$k</span>*2<span>]];
</span><span>54</span> <span> }
</span><span>55</span> <span> }
</span><span>56</span>
<span>57</span> <span> }
</span><span>58</span> <span>return</span> <span>$doubleArray</span><span>;
</span><span>59</span> }
五、倾斜调整
我们发现第三个9有一点倾斜，
我们需要将倾斜的图片“正”过来
人怎么处理的呢，先眼睛观察“倾斜了多少度”，然后把图片扭过来多少度，并且观察->负反馈->大脑传递扭转角度时刻在发生，最后图片就“正”过来了。
人是怎么观察“倾斜”的，以上面的“2”做例子，可能是右上方（左下方）的黑色比左上方（右下方）的多？
我们建立X轴正向向下，Y轴向右的直角坐标系
我们计算每一层的黑点的分布中点坐标，得到一系列离散点，计算这些点所在的直线（线性回归方程的计算，），公式y = b*x+a，
竟然有用到这个公式的一天！！！
大概就是一条倾斜的直线了，通过直线计算直线倾斜角度，然后转这么多的角度，图片应该就“正”了吧。
其中a,b的计算如下
<span> 1</span> <span>/*</span>
<span> 2</span> <span> *定义求线性回归A和B的函数
</span><span> 3</span> <span> *@param $zuobiaoArray坐标的三维数组
</span><span> 4</span> <span>*/</span>
<span> 5</span> <span>function</span> getHuigui(<span>$zuobiaoArray</span><span>){
</span><span> 6</span> <span>$y8</span> = 0<span>;
</span><span> 7</span> <span>$x8</span> = 0<span>;
</span><span> 8</span> <span>$x2</span> = 0<span>;
</span><span> 9</span> <span>$xy</span> = 0<span>;
</span><span>10</span> <span>$geshu</span> = <span>count</span>(<span>$zuobiaoArray</span><span>);
</span><span>11</span> <span>for</span>(<span>$i</span>=0;<span>$i</span><<span>$geshu</span>;<span>$i</span>++<span>){
</span><span>12</span> <span>$y8</span> = <span>$y8</span>+<span>$zuobiaoArray</span>[<span>$i</span>]['y'<span>];
</span><span>13</span> <span>$x8</span> = <span>$x8</span>+<span>$zuobiaoArray</span>[<span>$i</span>]['x'<span>];
</span><span>14</span> <span>$xy</span> = <span>$xy</span>+<span>$zuobiaoArray</span>[<span>$i</span>]['y']*<span>$zuobiaoArray</span>[<span>$i</span>]['x'<span>];
</span><span>15</span> <span>$x2</span> = <span>$x2</span> + <span>$zuobiaoArray</span>[<span>$i</span>]['x']*<span>$zuobiaoArray</span>[<span>$i</span>]['x'<span>];;
</span><span>16</span> <span> }
</span><span>17</span> <span>$y8</span> = <span>$y8</span>/<span>$geshu</span><span>;
</span><span>18</span> <span>$x8</span> = <span>$x8</span>/<span>$geshu</span><span>;
</span><span>19</span>
<span>20</span> <span>$b</span> = (<span>$xy</span>-<span>$geshu</span>*<span>$y8</span>*<span>$x8</span>)/(<span>$x2</span>-<span>$geshu</span>*<span>$x8</span>*<span>$x8</span><span>);
</span><span>21</span> <span>$a</span> = <span>$y8</span>-<span>$b</span>*<span>$x8</span><span>;
</span><span>22</span> <span>$re</span>['a'] = <span>$a</span><span>;
</span><span>23</span> <span>$re</span>['b'] = <span>$b</span><span>;
</span><span>24</span> <span>return</span> <span>$re</span><span>;
</span><span>25</span> <span>//</span><span>y = b * x + a</span>
<span>26</span> }
怎么转角？
1、可以直接对图片进行操作，但是发现有比较大的失真，就没有继续了。
2、或者，对黑点白点的坐标进行操作……
这就是三角函数了，好长时间不碰三角函数，都差点忘记了。
定义函数
<span> 1</span> <span>/*</span>
<span> 2</span> <span> *定义转化坐标的函数
</span><span> 3</span> <span> *@param $x x坐标即$i
</span><span> 4</span> <span> *@param $y y坐标，即j
</span><span> 5</span> <span> *@param $b 线性回归方程的b参数
</span><span> 6</span> <span>*/</span>
<span> 7</span> <span>function</span> getNewZuobiao(<span>$x</span>,<span>$y</span>,<span>$b</span><span>){
</span><span> 8</span> <span>if</span>(<span>$x</span> == 0<span>){
</span><span> 9</span> <span>if</span>(<span>$y</span>>0<span>){
</span><span>10</span> <span>$xianJiao</span> = M_PI/2<span>;
</span><span>11</span> }<span>elseif</span>(<span>$y</span><0<span>){
</span><span>12</span> <span>$xianJiao</span> = -M_PI/2<span>;
</span><span>13</span> }<span>else</span><span>{
</span><span>14</span> <span>$p</span>['x'] = 0<span>;
</span><span>15</span> <span>$p</span>['y'] = 0<span>;
</span><span>16</span> <span>return</span> <span>$p</span><span>;
</span><span>17</span> <span> }
</span><span>18</span> }<span>else</span><span>{
</span><span>19</span> <span>$xianJiao</span> = <span>atan</span>(<span>$y</span>/<span>$x</span><span>);
</span><span>20</span> <span> }
</span><span>21</span> <span>$jiao</span> =<span>$xianJiao</span>-<span>atan</span>(<span>$b</span><span>);
</span><span>22</span> <span>$chang</span> = <span>sqrt</span>(<span>$x</span>*<span>$x</span>+<span>$y</span>*<span>$y</span><span>);
</span><span>23</span> <span>$p</span>['x'] = <span>$chang</span>*<span>cos</span>(<span>$jiao</span><span>);
</span><span>24</span> <span>$p</span>['y'] = <span>$chang</span>*<span>sin</span>(<span>$jiao</span><span>);
</span><span>25</span> <span>return</span> <span>$p</span><span>;
</span><span>26</span> }
转角吧
<span> 1</span> <span>/*</span>
<span> 2</span> <span> *对【单个】数字的二值化二维数组进行倾斜调整
</span><span> 3</span> <span> *@param $singleArray 高*宽的二值数组（0,1）
</span><span> 4</span> <span>*/</span>
<span> 5</span> <span>function</span> singleSlopeAdjust(<span>$singleErzhiArray</span><span>){
</span><span> 6</span> <span>$slopeArray</span> = <span>array</span><span>();
</span><span> 7</span> <span>$gao</span> = <span>count</span>(<span>$singleErzhiArray</span><span>);
</span><span> 8</span> <span>$chang</span> = <span>count</span>(<span>$singleErzhiArray</span>['0'<span>]);
</span><span> 9</span>
<span>10</span> <span>//</span><span>初始化$slopeArray</span>
<span>11</span> <span>for</span>(<span>$i</span>=0;<span>$i</span><<span>$gao</span>*4;<span>$i</span>++<span>){
</span><span>12</span> <span>for</span>(<span>$j</span>=0;<span>$j</span><<span>$chang</span>*4;<span>$j</span>++<span>){
</span><span>13</span> <span>$slopeArray</span>[<span>$i</span>][<span>$j</span>] = 0<span>;
</span><span>14</span> <span> }
</span><span>15</span> <span> }
</span><span>16</span>
<span>17</span> <span>//</span><span>初始化中心坐标(是数组的下标)</span>
<span>18</span> <span>$centerXfoalt</span> = (<span>$gao</span>-1)/2<span>;
</span><span>19</span> <span>$centerYfoalt</span> = (<span>$chang</span>-1)/2<span>;
</span><span>20</span> <span>$centerX</span> = <span>ceil</span>(<span>$centerXfoalt</span><span>);
</span><span>21</span> <span>$centerY</span> = <span>ceil</span>(<span>$centerYfoalt</span><span>);
</span><span>22</span>
<span>23</span> <span>//</span><span>初始化图片倾斜诶角度</span>
<span>24</span> <span>/*</span><span>斜率的计算！！！！！，回归方程</span><span>*/</span>
<span>25</span> <span>//</span><span>从上到下扫描，计算中点，求得一串坐标（$i,$ava）</span>
<span>26</span> <span>for</span>(<span>$i</span>=0;<span>$i</span><<span>$gao</span>;<span>$i</span>++<span>){
</span><span>27</span> <span>$Num</span> = 0<span>;
</span><span>28</span> <span>$Amount</span> = 0<span>;
</span><span>29</span> <span>for</span>(<span>$j</span>=0;<span>$j</span><<span>$chang</span>;<span>$j</span>++<span>){
</span><span>30</span> <span>if</span>(<span>$singleErzhiArray</span>[<span>$i</span>][<span>$j</span>] == 1<span>){
</span><span>31</span> <span>$Num</span> = <span>$Num</span>+<span>$j</span><span>;
</span><span>32</span> <span>$Amount</span>++<span>;
</span><span>33</span> <span> }
</span><span>34</span> <span> }
</span><span>35</span> <span>if</span>(<span>$Amount</span> == 0<span>){
</span><span>36</span> <span>$Ava</span>[<span>$i</span>] = <span>$chang</span>/2<span>;
</span><span>37</span> }<span>else</span><span>{
</span><span>38</span> <span>$Ava</span>[<span>$i</span>] = <span>$Num</span>/<span>$Amount</span><span>;
</span><span>39</span> <span> }
</span><span>40</span> <span> }
</span><span>41</span>
<span>42</span>
<span>43</span> <span>//</span><span>计算线性回归方程的b与a</span>
<span>44</span> <span>$zuo</span> = <span>array</span><span>();
</span><span>45</span> <span>for</span>(<span>$j</span>=0;<span>$j</span><<span>count</span>(<span>$Ava</span>);<span>$j</span>++<span>){
</span><span>46</span> <span>$zuo</span>[<span>$j</span>]['x'] = <span>$j</span><span>;
</span><span>47</span> <span>$zuo</span>[<span>$j</span>]['y'] = <span>$Ava</span>[<span>$j</span><span>];
</span><span>48</span> <span> }
</span><span>49</span> <span>$res</span> = <span>$this</span>->getHuigui(<span>$zuo</span><span>);
</span><span>50</span> <span>$zuoB</span> = <span>$res</span>['b'<span>];
</span><span>51</span>
<span>52</span>
<span>53</span> <span>for</span>(<span>$i</span>=0;<span>$i</span><<span>$gao</span>;<span>$i</span>++<span>){
</span><span>54</span> <span>for</span>(<span>$j</span>=0;<span>$j</span><<span>$chang</span>;<span>$j</span>++<span>){
</span><span>55</span> <span>if</span>(<span>$singleErzhiArray</span>[<span>$i</span>][<span>$j</span>] == 1<span>){
</span><span>56</span> <span>$splodeZuobiao</span> = <span>$this</span>->getNewZuobiao(<span>$i</span>,<span>$j</span>,<span>$zuoB</span><span>);
</span><span>57</span> <span>$splodeX</span> = <span>$splodeZuobiao</span>['x'<span>];
</span><span>58</span> <span>$splodeY</span> = <span>$splodeZuobiao</span>['y'<span>];
</span><span>59</span> <span>$slopeArray</span>[<span>$splodeX</span>+<span>$gao</span>][<span>$splodeY</span>+<span>$chang</span>] = 1<span>;
</span><span>60</span> <span> }
</span><span>61</span> <span> }
</span><span>62</span> <span> }
</span><span>63</span>
<span>64</span> <span>//</span><span>将预处理的数组空白清理</span>
<span>65</span> <span>$slopeArray</span> = <span>$this</span>->getJinsuo(<span>$slopeArray</span><span>);
</span><span>66</span> <span>return</span> <span>$slopeArray</span><span>;
</span><span>67</span> }
看到正了一些
六、统一大小
上文中因为各种操作，每个字符大小不一，我们需要统一大小
七、特征值的建立
有很多方法
1、逐像素特征提取法
这是一种最简单的特征提取方法。它可以对图像进行逐行逐列的扫描，当遇到黑色像素时取其特征值为1，遇到白色像素时取其特征值为0，这样当扫描结束后就获得一个维数与图像中的像素点的个数相同的特征向量矩阵。
这种方法提取的信息量最大，但是它的缺点也很明显，就是适应性不强。
2、骨架特征提取法
两幅图像由于它们的线条的粗细不同，使得两幅图像差别很大，但是将它们的线条进行细化后，统一到相同的宽度，如一个像素宽时，这是两幅图像的差距就不那么明显。利用图形的骨架作为特征来进行数码识别，就使得识别有了一定的适应性。一般使用细化的方法来提取骨架，细化的算法有很多，如Hilditch算法、Rosenfeld算法等。对经过细化的图像利用EveryPixel函数进行处理就可以得到细化后图像的特征向量矩阵。骨架特征提取的方法对于线条粗细不同的数码有一定的适应性，但是图像一旦出现偏移就难以识别。
3、微结构法
微结构法将图像分为几个小块，统计每个小块的像素分布。本文提取出汉字的39个特征，存储在数组f[0]～f[38]中。具体算法可分为四步：
步骤一：把字符平均分成9份，如图4.1所示，给每一份编号如图4.2，统计每一份内黑色像素的个数，存储在数字tz[0]～tz[9]中，统计在行方向和列方向上每一份内的黑色像素个数和与之相邻的一份内黑色像素个数的比值作为一个特征，例如：行方向上提取特征f[0]＝tz[1]/ tz[0]，f[1]＝tz[2]/ tz[1]，f[2]＝tz[0]/ tz[2]，…,f[8]＝tz[6]/ tz[8]；列方向上f[9]＝tz[3]/ tz[0]，f[10]＝tz[6]/ tz[3]，f[11]＝tz[0]/ tz[6]，…，f[17]＝tz[2]/ tz[8]，共18个特征。
步骤二：把字符横向分成三份，如图4.3所示，统计每一份内的黑色像素个数，每一份内的黑色像素个数与前一份内黑色像素个数的比值作为一个特征，f[18]＝tz[10]/ tz[9]，f[19]＝tz[11]/ tz[10]，f[20]＝tz[9]/ tz[11]；把字符纵向分成三份，如图4.4所示，统计每一份内的黑色像素个数，每一份内的黑色像素个数与前一份内黑色像素个数的比值作为一个特征，f[21]＝tz[13]/ tz[12]，f[22]＝tz[14]/ tz[13]，f[23]＝tz[12]/ tz[14]；共六个特征。
步骤三：如图4.5，在竖直方向上找出三列，统计在该列中跳变点的个数，即相邻点像素值从0变到255的次数，共三个特征，记为f[24]，f[25]，f[26]；在水平方向上找出三行列，统计在该行中跳变点的个数，即相邻点象素值从0变到255的次数，共三个特征，记为f[27]，f[28]，f[29]。
图4.5
步骤四：把每一份内黑色象素的个数tz[0]～tz[9]，作为9个特征，记为：f[30]～f[38]。
这样得到汉字的共39个特征，根据这些特征就可以区分每个车牌汉字，进行识别。
我们使用最简单的逐像素特征提取法。
多多增加数据库，识别率会增加的
八、识别验证码
对于一个新的验证码，进行上文操作，然后对比数据库就可以了
<span> 1</span> <span>/*</span>
<span> 2</span> <span> *进行匹配
</span><span> 3</span> <span> *@param $Image 图片路径
</span><span> 4</span> <span>*/</span>
<span> 5</span> <span>public</span> <span>function</span> run(<span>$Image</span><span>){
</span><span> 6</span> <span>$data</span> = <span>array</span>('','','',''<span>);
</span><span> 7</span> <span>$result</span>=""<span>;
</span><span> 8</span> <span>$bilu</span> = ''<span>;
</span><span> 9</span> <span>$maxarr</span> = ''<span>;
</span><span>10</span>
<span>11</span> <span>//</span><span>提取特征</span>
<span>12</span> <span>$this</span>->prepare(<span>$Image</span><span>);
</span><span>13</span> <span>$yuanshi</span> = <span>$this</span>-><span>getErzhi();
</span><span>14</span> <span>$yijijiangzao</span> = <span>$this</span>->reduceZao(<span>$yuanshi</span><span>);
</span><span>15</span> <span>$small</span> = <span>$this</span>->cutSmall(<span>$yijijiangzao</span><span>);
</span><span>16</span> <span>for</span>(<span>$k</span>=0;<span>$k</span><4;<span>$k</span>++<span>){
</span><span>17</span> <span>$tianchong</span> = <span>$this</span>->tianChong(<span>$small</span>[<span>$k</span><span>]);
</span><span>18</span> <span>$tiaozhenjiaodu</span> = <span>$this</span>->singleSlopeAdjust(<span>$tianchong</span><span>);
</span><span>19</span> <span>$tongyidaxiao</span> = <span>$this</span>->tongyiDaxiao(<span>$tiaozhenjiaodu</span><span>);
</span><span>20</span> <span>for</span>(<span>$i</span>=0;<span>$i</span><20;<span>$i</span>++<span>){
</span><span>21</span> <span>for</span>(<span>$j</span>=0;<span>$j</span><20;<span>$j</span>++<span>){
</span><span>22</span> <span>$data</span>[<span>$k</span>] .= <span>$tongyidaxiao</span>[<span>$i</span>][<span>$j</span><span>];
</span><span>23</span> <span> }
</span><span>24</span> <span> }
</span><span>25</span> <span> }
</span><span>26</span>
<span>27</span> <span>//</span><span> 进行关键字匹配</span>
<span>28</span> <span>foreach</span>(<span>$data</span> <span>as</span> <span>$numKey</span> => <span>$numString</span><span>)
</span><span>29</span> <span> {
</span><span>30</span>
<span>31</span> <span>$max</span> = 0<span>;
</span><span>32</span> <span>$num</span> = 0<span>;
</span><span>33</span> <span>foreach</span>(<span>$this</span>->Keys <span>as</span> <span>$value</span> => <span>$key</span><span>)
</span><span>34</span> <span> {
</span><span>35</span> <span>similar_text</span>(<span>$value</span>, <span>$numString</span>,<span>$percent</span><span>);
</span><span>36</span> <span>if</span>(<span>$percent</span> > <span>$max</span><span>)
</span><span>37</span> <span> {
</span><span>38</span> <span>$max</span> = <span>$percent</span><span>;
</span><span>39</span> <span>$num</span> = <span>$key</span><span>;
</span><span>40</span> <span>$zim</span> = <span>$value</span><span>;
</span><span>41</span> <span> }
</span><span>42</span> <span>if</span>(<span>$max</span>>95<span>){
</span><span>43</span> <span>break</span><span>;
</span><span>44</span> <span> }
</span><span>45</span> <span> }
</span><span>46</span> <span>$result</span> .=<span>$num</span><span>;
</span><span>47</span> <span>$maxarr</span>[] = <span>$max</span><span>;
</span><span>48</span> <span> }
</span><span>49</span> <span>//</span><span> 查找最佳匹配数字</span>
<span>50</span> <span>$re</span> = <span>$maxarr</span><span>;
</span><span>51</span> <span>$re</span>[] = <span>$result</span><span>;
</span><span>52</span> <span>return</span> <span>$re</span><span>;
</span><span>53</span> <span>//</span><span>return $result.'|max|一:'.$maxarr['0'].'|二:'.$maxarr['1'].'|三:'.$maxarr['2'].'|四:'.$maxarr['3'];</span>
<span>54</span> }
试试：
Header("Content-type: image/gif");
/*
* 初始化
*/
$border = 0; //是否要边框 1要:0不要
$how = 4; //验证码位数
$w = $how*15; //图片宽度
$h = 20; //图片高度
$fontsize = 5; //字体大小
$alpha = "abcdefghijkmnopqrstuvwxyz"; //验证码内容1:字母
$number = "023456789"; //验证码内容2:数字
$randcode = ""; //验证码字符串初始化
srand((double)microtime()*1000000); //初始化随机数种子
$im = ImageCreate($w, $h); //创建验证图片
/*
* 绘制基本框架
*/
$bgcolor = ImageColorAllocate($im, 255, 255, 255); //设置背景颜色
ImageFill($im, 0, 0, $bgcolor); //填充背景色
if($border)
{
$black = ImageColorAllocate($im, 0, 0, 0); //设置边框颜色
ImageRectangle($im, 0, 0, $w-1, $h-1, $black);//绘制边框
}
/*
* 逐位产生随机字符
*/
for($i=0; $i{
$alpha_or_number = mt_rand(0, 1); //字母还是数字
$str = $alpha_or_number ? $alpha : $number;
$which = mt_rand(0, strlen($str)-1); //取哪个字符
$code = substr($str, $which, 1); //取字符
$j = !$i ? 4 : $j+15; //绘字符位置
$color3 = ImageColorAllocate($im, mt_rand(0,100), mt_rand(0,100), mt_rand(0,100)); //字符随即颜色
ImageChar($im, $fontsize, $j, 3, $code, $color3); //绘字符
$randcode .= $code; //逐位加入验证码字符串
}
/*
* 添加干扰
*/
for($i=0; $i{
$color1 = ImageColorAllocate($im, mt_rand(0,255), mt_rand(0,255), mt_rand(0,255)); //干扰线颜色
ImageArc($im, mt_ra......余下全文>>
原理不难，只需通过PHP来生成一张图片！在这张图片上通过PHP的图片处理函数在图片上写几个字符就Ok啦！在写字符时注意将写的字符记录下，这个记录是通过session来做！然后在验证时直接调用$_SESSION变量就可以了！
具体的代码如下：
有两个文件
yangzhengma.php用来显示图片和验证输入的验证码
session_start();
if(!empty($_GET['sub'])){
if($_SESSION['checks'] == $_GET['ma']){
echo "验证码正确
";
}else{
echo "验证码不正确！
";
}
}
?>
![[Web front-end] Permulaan pantas Node.js](https://img.php.cn/upload/course/000/000/067/662b5d34ba7c0227.png)
