java验证码识别--1 收藏

(本文仅用于学习研究图像匹配识别原理，不得用于其他用途。

 

最近看了看验证码的识别，先从最简单的做起吧(固定大小，固定位置，固定字体)

验证码识别基本分四步，图片预处理，分割，训练，识别

看一个最简单验证码

 

这是一个德克萨斯扑克的注册页面的验证码

1。图像的预处理

这种直接根据亮度设个阈值处理就可以了

view plaincopy to clipboardprint?
public static int isWhite(int colorInt) {  
        Color color = new Color(colorInt);  
        if (color.getRed() + color.getGreen() + color.getBlue() > 100) {  
            return 1;  
        }  
        return 0;  
    }  
 
    public static BufferedImage removeBackgroud(String picFile)  
            throws Exception {  
        BufferedImage img = ImageIO.read(new File(picFile));  
        int width = img.getWidth();  
        int height = img.getHeight();  
        for (int x = 0; x < width; ++x) {  
            for (int y = 0; y < height; ++y) {  
                if (isWhite(img.getRGB(x, y)) == 1) {  
                    img.setRGB(x, y, Color.WHITE.getRGB());  
                } else {  
                    img.setRGB(x, y, Color.BLACK.getRGB());  
                }  
            }  
        }  
        return img;  
    } 
public static int isWhite(int colorInt) {
  Color color = new Color(colorInt);
  if (color.getRed() + color.getGreen() + color.getBlue() > 100) {
   return 1;
  }
  return 0;
 }

 public static BufferedImage removeBackgroud(String picFile)
   throws Exception {
  BufferedImage img = ImageIO.read(new File(picFile));
  int width = img.getWidth();
  int height = img.getHeight();
  for (int x = 0; x < width; ++x) {
   for (int y = 0; y < height; ++y) {
    if (isWhite(img.getRGB(x, y)) == 1) {
     img.setRGB(x, y, Color.WHITE.getRGB());
    } else {
     img.setRGB(x, y, Color.BLACK.getRGB());
    }
   }
  }
  return img;
 }

处理完图片效果为

 

图像基本分得比较清楚，图片分割也比较容易

2。分割

这个验证码居然是固定位置的，分割相当简单，直接截取相应位置就可以了

view plaincopy to clipboardprint?
public static List<BufferedImage> splitImage(BufferedImage img)  
            throws Exception {  
        List<BufferedImage> subImgs = new ArrayList<BufferedImage>();  
        subImgs.add(img.getSubimage(10, 6, 8, 10));  
        subImgs.add(img.getSubimage(19, 6, 8, 10));  
        subImgs.add(img.getSubimage(28, 6, 8, 10));  
        subImgs.add(img.getSubimage(37, 6, 8, 10));  
        return subImgs;  
    } 
public static List<BufferedImage> splitImage(BufferedImage img)
   throws Exception {
  List<BufferedImage> subImgs = new ArrayList<BufferedImage>();
  subImgs.add(img.getSubimage(10, 6, 8, 10));
  subImgs.add(img.getSubimage(19, 6, 8, 10));
  subImgs.add(img.getSubimage(28, 6, 8, 10));
  subImgs.add(img.getSubimage(37, 6, 8, 10));
  return subImgs;
 }

3。训练

直接拿几张图片，包含0-9，每个数字一个样本就可以了，将文件名对应相应的数字

 

4。识别

因为是固定大小，固定位置，识别也很简单。

直接拿分割的图片跟这个十个图片一个像素一个像素的比，相同的点最多的就是结果。比如如果跟5.jpg最相似，那么识别的结果就是5。

下面是识别结果，很容易达到100%

 

完整代码(csdn连个附件都不支持)：

view plaincopy to clipboardprint?
import java.awt.Color;  
import java.awt.image.BufferedImage;  
import java.io.File;  
import java.io.FileOutputStream;  
import java.io.InputStream;  
import java.io.OutputStream;  
import java.util.ArrayList;  
import java.util.HashMap;  
import java.util.List;  
import java.util.Map;  
 
import javax.imageio.ImageIO;  
 
import org.apache.commons.httpclient.HttpClient;  
import org.apache.commons.httpclient.HttpStatus;  
import org.apache.commons.httpclient.methods.GetMethod;  
import org.apache.commons.io.IOUtils;  
 
public class ImagePreProcess {  
 
    public static int isWhite(int colorInt) {  
        Color color = new Color(colorInt);  
        if (color.getRed() + color.getGreen() + color.getBlue() > 100) {  
            return 1;  
        }  
        return 0;  
    }  
 
    public static int isBlack(int colorInt) {  
        Color color = new Color(colorInt);  
        if (color.getRed() + color.getGreen() + color.getBlue() <= 100) {  
            return 1;  
        }  
        return 0;  
    }  
 
    public static BufferedImage removeBackgroud(String picFile)  
            throws Exception {  
        BufferedImage img = ImageIO.read(new File(picFile));  
        int width = img.getWidth();  
        int height = img.getHeight();  
        for (int x = 0; x < width; ++x) {  
            for (int y = 0; y < height; ++y) {  
                if (isWhite(img.getRGB(x, y)) == 1) {  
                    img.setRGB(x, y, Color.WHITE.getRGB());  
                } else {  
                    img.setRGB(x, y, Color.BLACK.getRGB());  
                }  
            }  
        }  
        return img;  
    }  
 
    public static List<BufferedImage> splitImage(BufferedImage img)  
            throws Exception {  
        List<BufferedImage> subImgs = new ArrayList<BufferedImage>();  
        subImgs.add(img.getSubimage(10, 6, 8, 10));  
        subImgs.add(img.getSubimage(19, 6, 8, 10));  
        subImgs.add(img.getSubimage(28, 6, 8, 10));  
        subImgs.add(img.getSubimage(37, 6, 8, 10));  
        return subImgs;  
    }  
 
    public static Map<BufferedImage, String> loadTrainData() throws Exception {  
        Map<BufferedImage, String> map = new HashMap<BufferedImage, String>();  
        File dir = new File("train");  
        File[] files = dir.listFiles();  
        for (File file : files) {  
            map.put(ImageIO.read(file), file.getName().charAt(0) + "");  
        }  
        return map;  
    }  
 
    public static String getSingleCharOcr(BufferedImage img,  
            Map<BufferedImage, String> map) {  
        String result = "";  
        int width = img.getWidth();  
        int height = img.getHeight();  
        int min = width * height;  
        for (BufferedImage bi : map.keySet()) {  
            int count = 0;  
            Label1: for (int x = 0; x < width; ++x) {  
                for (int y = 0; y < height; ++y) {  
                    if (isWhite(img.getRGB(x, y)) != isWhite(bi.getRGB(x, y))) {  
                        count++;  
                        if (count >= min)  
                            break Label1;  
                    }  
                }  
            }  
            if (count < min) {  
                min = count;  
                result = map.get(bi);  
            }  
        }  
        return result;  
    }  
 
    public static String getAllOcr(String file) throws Exception {  
        BufferedImage img = removeBackgroud(file);  
        List<BufferedImage> listImg = splitImage(img);  
        Map<BufferedImage, String> map = loadTrainData();  
        String result = "";  
        for (BufferedImage bi : listImg) {  
            result += getSingleCharOcr(bi, map);  
        }  
        ImageIO.write(img, "JPG", new File("result\\"+result+".jpg"));  
        return result;  
    }  
 
    public static void downloadImage() {  
        HttpClient httpClient = new HttpClient();  
        GetMethod getMethod = new GetMethod(  
                "http://www.puke888.com/authimg.php");  
        for (int i = 0; i < 30; i++) {  
            try {  
                // 执行getMethod  
                int statusCode = httpClient.executeMethod(getMethod);  
                if (statusCode != HttpStatus.SC_OK) {  
                    System.err.println("Method failed: " 
                            + getMethod.getStatusLine());  
                }  
                // 读取内容  
                String picName = "img\\" + i + ".jpg";  
                InputStream inputStream = getMethod.getResponseBodyAsStream();  
                OutputStream outStream = new FileOutputStream(picName);  
                IOUtils.copy(inputStream, outStream);  
                outStream.close();  
                System.out.println("OK!");  
            } catch (Exception e) {  
                e.printStackTrace();  
            } finally {  
                // 释放连接  
                getMethod.releaseConnection();  
            }  
        }  
    }  
 
    /** 
     * @param args 
     * @throws Exception 
     */ 
    public static void main(String[] args) throws Exception {  
        for (int i = 0; i < 30; ++i) {  
            String text = getAllOcr("img\\" + i + ".jpg");  
            System.out.println(i + ".jpg = " + text);  
        }  
    }  
} 

 

本文来自CSDN博客，转载请标明出处：http://blog.csdn.net/problc/archive/2010/08/07/5794460.aspx

 

 

 

个人签名

-------------------------------------

 

图盾 淘宝保护 保护图片 图片防盗