Java OCR tesseract 图像智能字符识别技术 Java代码实现

分享一下我老师大神的人工智能教程。零基础!通俗易懂!风趣幽默!还带黄段子!希望你也加入到我们人工智能的队伍中来!https://blog.csdn.net/jiangjunshow

               

接着上一篇OCR所说的,上一篇给大家介绍了tesseract 在命令行的简单用法,当然了要继承到我们的程序中,还是需要代码实现的,下面给大家分享下java实现的例子。


拿代码扫描上面的图片,然后输出结果。主要思想就是利用Java调用系统任务。

下面是核心代码:

package com.zhy.test;import java.io.BufferedReader;import java.io.File;import java.io.FileInputStream;import java.io.InputStreamReader;import java.util.ArrayList;import java.util.List;import org.jdesktop.swingx.util.OS;public class OCRHelperprivate final String LANG_OPTION = "-l"private final String EOL = System.getProperty("line.separator"); /**  * 文件位置我防止在,项目同一路径  */ private String tessPath = new File("tesseract").getAbsolutePath(); /**  * @param imageFile  *            传入的图像文件  * @param imageFormat  *            传入的图像格式  * @return 识别后的字符串  */ public String recognizeText(File imageFile) throws Exception {  /**   * 设置输出文件的保存的文件目录   */  File outputFile = new File(imageFile.getParentFile(), "output");  StringBuffer strB = new StringBuffer();  List cmd = new ArrayList();  if (OS.isWindowsXP())  {   cmd.add(tessPath + "\\tesseract");  } else if (OS.isLinux())  {   cmd.add("tesseract");  } else  {   cmd.add(tessPath + "\\tesseract");  }  cmd.add("");  cmd.add(outputFile.getName());  cmd.add(LANG_OPTION);//  cmd.add("chi_sim");  cmd.add("eng");  ProcessBuilder pb = new ProcessBuilder();  /**   *Sets this process builder's working directory.   */  pb.directory(imageFile.getParentFile());  cmd.set(1, imageFile.getName());  pb.command(cmd);  pb.redirectErrorStream(true);  Process process = pb.start();  // tesseract.exe 1.jpg 1 -l chi_sim  // Runtime.getRuntime().exec("tesseract.exe 1.jpg 1 -l chi_sim");  /**   * the exit value of the process. By convention, 0 indicates normal   * termination.   *///  System.out.println(cmd.toString());  int w = process.waitFor();  if (w == 0)// 0代表正常退出  {   BufferedReader in = new BufferedReader(new InputStreamReader(     new FileInputStream(outputFile.getAbsolutePath() + ".txt"),     "UTF-8"));   String str;   while ((str = in.readLine()) != null)   {    strB.append(str).append(EOL);   }   in.close();  } else  {   String msg;   switch (w)   {   case 1:    msg = "Errors accessing files. There may be spaces in your image's filename.";    break;   case 29:    msg = "Cannot recognize the image or its selected region.";    break;   case 31:    msg = "Unsupported image format.";    break;   default:    msg = "Errors occurred.";   }   throw new RuntimeException(msg);  }  new File(outputFile.getAbsolutePath() + ".txt").delete();  return strB.toString().replaceAll("\\s*", ""); }}
代码很简单,中间那部分ProcessBuilder其实就类似Runtime.getRuntime().exec("tesseract.exe 1.jpg 1 -l chi_sim"),大家不习惯的可以使用Runtime。

测试代码:

package com.zhy.test;import java.io.File;public class Testpublic static void main(String[] args) {  try  {      File testDataDir = new File("testdata");   System.out.println(testDataDir.listFiles().length);   int i = 0 ;    for(File file :testDataDir.listFiles())   {    i++ ;    String recognizeText = new OCRHelper().recognizeText(file);    System.out.print(recognizeText+"\t");    if( i % 5  == 0 )    {     System.out.println();    }   }     } catch (Exception e)  {   e.printStackTrace();  } }}

输出结果:


对比第一张图片,是不是很完美~哈哈 ,当然了如果你只需要实现验证码的读写,那么上面就足够了。下面继续普及图像处理的知识。



-------------------------------------------------------------------我的分割线--------------------------------------------------------------------

当然了,有时候图片被扭曲或者模糊的很厉害,很不容易识别,所以下面我给大家介绍一个去噪的辅助类,绝对碉堡了,先看下效果图。


来张特写:


一个类,不依赖任何jar,把图像中的干扰线消灭了,是不是很给力,然后再拿这样的图片去识别,会不会效果更好呢,嘿嘿,大家自己实验~

代码:

package com.zhy.test;import java.awt.Color;import java.awt.image.BufferedImage;import java.io.File;import java.io.IOException;import javax.imageio.ImageIO;public class ClearImageHelperpublic static void main(String[] args) throws IOException {    File testDataDir = new File("testdata");  final String destDir = testDataDir.getAbsolutePath()+"/tmp";  for (File file : testDataDir.listFiles())  {   cleanImage(file, destDir);  } } /**  *   * @param sfile  *            需要去噪的图像  * @param destDir  *            去噪后的图像保存地址  * @throws IOException  */ public static void cleanImage(File sfile, String destDir)   throws IOException {  File destF = new File(destDir);  if (!destF.exists())  {   destF.mkdirs();  }  BufferedImage bufferedImage = ImageIO.read(sfile);  int h = bufferedImage.getHeight();  int w = bufferedImage.getWidth();  // 灰度化  int[][] gray = new int[w][h];  for (int x = 0; x < w; x++)  {   for (int y = 0; y < h; y++)   {    int argb = bufferedImage.getRGB(x, y);    // 图像加亮(调整亮度识别率非常高)    int r = (int) (((argb >> 16) & 0xFF) * 1.1 + 30);    int g = (int) (((argb >> 8) & 0xFF) * 1.1 + 30);    int b = (int) (((argb >> 0) & 0xFF) * 1.1 + 30);    if (r >= 255)    {     r = 255;    }    if (g >= 255)    {     g = 255;    }    if (b >= 255)    {     b = 255;    }    gray[x][y] = (int) Math      .pow((Math.pow(r, 2.2) * 0.2973 + Math.pow(g, 2.2)        * 0.6274 + Math.pow(b, 2.2) * 0.0753), 1 / 2.2);   }  }  // 二值化  int threshold = ostu(gray, w, h);  BufferedImage binaryBufferedImage = new BufferedImage(w, h,    BufferedImage.TYPE_BYTE_BINARY);  for (int x = 0; x < w; x++)  {   for (int y = 0; y < h; y++)   {    if (gray[x][y] > threshold)    {     gray[x][y] |= 0x00FFFF;    } else    {     gray[x][y] &= 0xFF0000;    }    binaryBufferedImage.setRGB(x, y, gray[x][y]);   }  }  // 矩阵打印  for (int y = 0; y < h; y++)  {   for (int x = 0; x < w; x++)   {    if (isBlack(binaryBufferedImage.getRGB(x, y)))    {     System.out.print("*");    } else    {     System.out.print(" ");    }   }   System.out.println();  }  ImageIO.write(binaryBufferedImage, "jpg", new File(destDir, sfile    .getName())); } public static boolean isBlack(int colorInt) {  Color color = new Color(colorInt);  if (color.getRed() + color.getGreen() + color.getBlue() <= 300)  {   return true;  }  return false; } public static boolean isWhite(int colorInt) {  Color color = new Color(colorInt);  if (color.getRed() + color.getGreen() + color.getBlue() > 300)  {   return true;  }  return false; } public static int isBlackOrWhite(int colorInt) {  if (getColorBright(colorInt) < 30 || getColorBright(colorInt) > 730)  {   return 1;  }  return 0; } public static int getColorBright(int colorInt) {  Color color = new Color(colorInt);  return color.getRed() + color.getGreen() + color.getBlue(); } public static int ostu(int[][] gray, int w, int h) {  int[] histData = new int[w * h];  // Calculate histogram  for (int x = 0; x < w; x++)  {   for (int y = 0; y < h; y++)   {    int red = 0xFF & gray[x][y];    histData[red]++;   }  }  // Total number of pixels  int total = w * h;  float sum = 0;  for (int t = 0; t < 256; t++)   sum += t * histData[t];  float sumB = 0;  int wB = 0;  int wF = 0;  float varMax = 0;  int threshold = 0;  for (int t = 0; t < 256; t++)  {   wB += histData[t]; // Weight Background   if (wB == 0)    continue;   wF = total - wB; // Weight Foreground   if (wF == 0)    break;   sumB += (float) (t * histData[t]);   float mB = sumB / wB; // Mean Background   float mF = (sum - sumB) / wF; // Mean Foreground   // Calculate Between Class Variance   float varBetween = (float) wB * (float) wF * (mB - mF) * (mB - mF);   // Check if new maximum found   if (varBetween > varMax)   {    varMax = varBetween;    threshold = t;   }  }  return threshold; }}


好了,就到这里。如果这篇文章对你有用,赞一个吧~





           

分享一下我老师大神的人工智能教程。零基础!通俗易懂!风趣幽默!还带黄段子!希望你也加入到我们人工智能的队伍中来!https://blog.csdn.net/jiangjunshow

你可能感兴趣的:(Java OCR tesseract 图像智能字符识别技术 Java代码实现)