驗證碼識別總結

     因爲公司的需求，這幾天研究下了驗證碼識別。對驗證碼識別大體分這幾個過程，第一步獲取驗證碼，第二對驗證碼處理，若是顏色單一沒什麼背景雜色就直接二值化處理，注意闕值，有干擾線的把干擾線和背景去掉，最終變爲背景爲白色，驗證碼前景色爲黑色。第三步就是切割，把驗證碼從圖片中切割出來，第四創建識別庫，切割後的圖片分類存入識別庫，讓後須要讓程序學習一些驗證碼後，識別庫就有了樣例。第四步就是那當前是別的驗證碼和識別庫的驗證碼進行比對，達到識別驗證碼的結果。

    識別類源碼：

using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Drawing;
using System.Net;
using System.Collections;
using System.IO;

namespace CheckCodeRecognizeLib
{
    public class CheckCodeRecognize
    {
        #region 成員變量
        //色差閥值 越小消除的雜色越多
        private double threshold = 150;
        //二值閥值 越大效果越不明顯
        private double ezFZ = 0.6;
        //背景近似度閥值
        private double bjfz = 80;
        //圖片路徑
        private string imgPath = string.Empty;
        //每一個字符最小寬度
        public int MinWidthPerChar = 7;
        //每一個字符最大寬度
        public int MaxWidthPerChar = 18;
        //每一個字符最小高度
        public int MinHeightPerChar = 10;
        //學習庫保存的路徑
        private readonly string samplePath = AppDomain.CurrentDomain.BaseDirectory + "Sample\\";
        #endregion

        #region 圖片處理
        /// <summary>
        /// 對傳入的圖片二值化
        /// </summary>
        /// <param name="bitmap">傳入的原圖片</param>
        /// <returns>處理事後的圖片</returns>
        private Bitmap EZH(Bitmap bitmap)
        {
            if (bitmap != null)
            {
                var img = new Bitmap(bitmap);
                for (var x = 0; x < img.Width; x++)
                {
                    for (var y = 0; y < img.Height; y++)
                    {
                        Color color = img.GetPixel(x, y);
                        if (color.GetBrightness() < ezFZ)
                        {
                            img.SetPixel(x, y, Color.Black);
                        }
                        else
                        {
                            img.SetPixel(x, y, Color.White);
                        }
                    }
                }
                return img;                
            }
            return null;
            
        }
        /// <summary>
        /// 去背景
        /// 把圖片中最多的一部分顏色視爲背景色 選出來後替換爲白色
        /// </summary>
        /// <param name="bitmapImg">將要處理的圖片</param>
        /// <returns>返回去過背景的圖片</returns>
        private Bitmap RemoveBackGround(Bitmap bitmapImg)
        {
            if (bitmapImg == null)
            {
                return null;
            }          
            //key 顏色  value顏色對應的數量
            Dictionary<Color, int> colorDic = new Dictionary<Color, int>();
            //獲取圖片中每一個顏色的數量
            for (var x = 0; x < bitmapImg.Width; x++)
            {
                for (var y = 0; y < bitmapImg.Height; y++)
                {
                    //刪除邊框
                    if (y == 0 || y == bitmapImg.Height)
                    {
                        bitmapImg.SetPixel(x, y, Color.White);
                    }

                    var color = bitmapImg.GetPixel(x, y);
                    var colorRGB = color.ToArgb();

                    if (colorDic.ContainsKey(color))
                    {
                        colorDic[color] = colorDic[color] + 1;
                    }
                    else
                    {
                        colorDic[color] = 1;
                    }
                }
            }
            //圖片中最多的顏色
            Color maxColor = colorDic.OrderByDescending(o => o.Value).FirstOrDefault().Key;
            //圖片中最少的顏色
            Color minColor = colorDic.OrderBy(o => o.Value).FirstOrDefault().Key;

            Dictionary<int[], double> maxColorDifDic = new Dictionary<int[], double>();
            //查找 maxColor 最接近顏色
            for (var x = 0; x < bitmapImg.Width; x++)
            {
                for (var y = 0; y < bitmapImg.Height; y++)
                {
                    maxColorDifDic.Add(new int[] { x, y }, GetColorDif(bitmapImg.GetPixel(x, y), maxColor));
                }
            }
            //去掉和maxColor接近的顏色 即 替換成白色
            var maxColorDifList = maxColorDifDic.OrderBy(o => o.Value).Where(o => o.Value < bjfz).ToArray();
            foreach (var kv in maxColorDifList)
            {
                bitmapImg.SetPixel(kv.Key[0], kv.Key[1], Color.White);
            }
            return bitmapImg;
           
        }
        /// <summary>
        /// 獲取色差
        /// </summary>
        /// <param name="color1"></param>
        /// <param name="color2"></param>
        /// <returns></returns>
        private double GetColorDif(Color color1, Color color2)
        {
            return Math.Sqrt((Math.Pow((color1.R - color2.R), 2) +
                Math.Pow((color1.G - color2.G), 2) +
                Math.Pow((color1.B - color2.B), 2)));
        }
        /// <summary>
        /// 去掉目標干擾線
        /// </summary>
        /// <param name="img">將要處理的圖片</param>
        /// <returns>去掉乾乾擾線處理過的圖片</returns>  
        private Bitmap btnDropDisturb_Click(Bitmap img)
        {
            if (img == null)
            {
                return null;
            }         
            byte[] p = new byte[9]; //最小處理窗口3*3
            //去幹擾線
            for (var x = 0; x < img.Width; x++)
            {
                for (var y = 0; y < img.Height; y++)
                {
                    Color currentColor = img.GetPixel(x, y);
                    int color = currentColor.ToArgb();

                    if (x > 0 && y > 0 && x < img.Width - 1 && y < img.Height - 1)
                    {
                        #region 中值濾波效果很差
                        ////取9個點的值
                        //p[0] = img.GetPixel(x - 1, y - 1).R;
                        //p[1] = img.GetPixel(x, y - 1).R;
                        //p[2] = img.GetPixel(x + 1, y - 1).R;
                        //p[3] = img.GetPixel(x - 1, y).R;
                        //p[4] = img.GetPixel(x, y).R;
                        //p[5] = img.GetPixel(x + 1, y).R;
                        //p[6] = img.GetPixel(x - 1, y + 1).R;
                        //p[7] = img.GetPixel(x, y + 1).R;
                        //p[8] = img.GetPixel(x + 1, y + 1).R;
                        ////計算中值
                        //for (int j = 0; j < 5; j++)
                        //{
                        //    for (int i = j + 1; i < 9; i++)
                        //    {
                        //        if (p[j] > p[i])
                        //        {
                        //            s = p[j];
                        //            p[j] = p[i];
                        //            p[i] = s;
                        //        }
                        //    }
                        //}
                        ////      if (img.GetPixel(x, y).R < dgGrayValue)
                        //img.SetPixel(x, y, Color.FromArgb(p[4], p[4], p[4]));    //給有效值付中值
                        #endregion

                        //上 x y+1
                        double upDif = GetColorDif(currentColor, img.GetPixel(x, y + 1));
                        //下 x y-1
                        double downDif = GetColorDif(currentColor, img.GetPixel(x, y - 1));
                        //左 x-1 y
                        double leftDif = GetColorDif(currentColor, img.GetPixel(x - 1, y));
                        //右 x+1 y
                        double rightDif = GetColorDif(currentColor, img.GetPixel(x + 1, y));
                        //左上
                        double upLeftDif = GetColorDif(currentColor, img.GetPixel(x - 1, y + 1));
                        //右上
                        double upRightDif = GetColorDif(currentColor, img.GetPixel(x + 1, y + 1));
                        //左下
                        double downLeftDif = GetColorDif(currentColor, img.GetPixel(x - 1, y - 1));
                        //右下
                        double downRightDif = GetColorDif(currentColor, img.GetPixel(x + 1, y - 1));

                        ////四面色差較大
                        //if (upDif > threshold && downDif > threshold && leftDif > threshold && rightDif > threshold)
                        //{
                        //    img.SetPixel(x, y, Color.White);
                        //}
                        //三面色差較大
                        if ((upDif > threshold && downDif > threshold && leftDif > threshold)
                            || (downDif > threshold && leftDif > threshold && rightDif > threshold)
                            || (upDif > threshold && leftDif > threshold && rightDif > threshold)
                            || (upDif > threshold && downDif > threshold && rightDif > threshold))
                        {
                            img.SetPixel(x, y, Color.White);
                        }

                        List<int[]> xLine = new List<int[]>();
                        //去橫向干擾線  原理 若是這個點上下有不少白色像素則認爲是干擾
                        for (var x1 = x + 1; x1 < x + 10; x1++)
                        {
                            if (x1 >= img.Width)
                            {
                                break;
                            }

                            if (img.GetPixel(x1, y + 1).ToArgb() == Color.White.ToArgb()
                                && img.GetPixel(x1, y - 1).ToArgb() == Color.White.ToArgb())
                            {
                                xLine.Add(new int[] { x1, y });
                            }
                        }
                        if (xLine.Count() >= 4)
                        {
                            foreach (var xpoint in xLine)
                            {
                                img.SetPixel(xpoint[0], xpoint[1], Color.White);
                            }
                        }

                        //去豎向干擾線

                    }
                }
            }
            return img;
        }
        /// <summary>
        /// 對圖片先豎向分割，再橫向分割
       /// </summary>
       /// <param name="img">將要分割的圖片</param>
        /// <returns>全部分割後的字符圖片</returns>
        private Bitmap[] SplitImage(Bitmap img)
        {
            if (img == null)
            {
                return null;
            }
            List<int[]> xCutPointList = GetXCutPointList(img);
            List<int[]> yCutPointList = GetYCutPointList(xCutPointList, img);       
            Bitmap[] bitmapArr = new Bitmap[5];
            //對分割的部分劃線
            for (int i = 0; i < xCutPointList.Count(); i++)
            {
                int xStart = xCutPointList[i][0];
                int xEnd = xCutPointList[i][1];
                int yStart = yCutPointList[i][0];
                int yEnd = yCutPointList[i][1];
                if (i >= 4) break;
                bitmapArr[i]= (Bitmap)AcquireRectangleImage(img,
                    new Rectangle(xStart, yStart, xEnd - xStart + 1, yEnd - yStart + 1));
            }
            return bitmapArr;
        }
        /// <summary>
        /// 分別從圖片的上下尋找像素點大於闕值的地方，而後獲取有黑色像素的有效區域
        /// </summary>
        /// <param name="xCutPointList">x軸範圍的x座標集合</param>
        /// <param name="img">目標圖片</param>
        /// <returns>y軸座標開始和結束點，其實就是黑色像素圖片的有效區域</returns>
        private List<int[]> GetYCutPointList(List<int[]> xCutPointList, Bitmap img)
        {
            List<int[]> list = new List<int[]>();
            //獲取圖像最上面Y值
            int topY = 0;
            //獲取圖像最下面的Y值
            int bottomY = 0;
            foreach (var xPoint in xCutPointList)
            {
                for (int ty = 1; ty < img.Height; ty++)
                {
                    int xStart = xPoint[0];
                    int xEnd = xPoint[1];
                    int blackCount = GetBlackPXCountInY(ty, 2, xStart, xEnd, img);
                    if (blackCount > 3)
                    {
                        topY = ty;
                        break;
                    }
                }
                for (int by = img.Height; by > 1; by--)
                {
                    int xStart = xPoint[0];
                    int xEnd = xPoint[1];
                    int blackCount = GetBlackPXCountInY(by, -2, xStart, xEnd, img);
                    if (blackCount > 3)
                    {
                        bottomY = by;
                        break;
                    }
                }
                list.Add(new int[] { topY, bottomY });

            }
            return list;
        }
        /// <summary>
        /// 獲取分割後某區域的黑色像素
        /// </summary>
        /// <param name="startY"></param>
        /// <param name="offset"></param>
        /// <param name="startX"></param>
        /// <param name="endX"></param>
        /// <param name="img"></param>
        /// <returns></returns>
        private int GetBlackPXCountInY(int startY, int offset, int startX, int endX, Bitmap img)
        {
            int blackPXCount = 0;
            int startY1 = offset > 0 ? startY : startY + offset;
            int offset1 = offset > 0 ? startY + offset : startY;
            for (var x = startX; x <= endX; x++)
            {
                for (var y = startY1; y < offset1; y++)
                {
                    if (y >= img.Height)
                    {
                        continue;
                    }
                    if (img.GetPixel(x, y).ToArgb() == Color.Black.ToArgb())
                    {
                        blackPXCount++;
                    }
                }
            }
            return blackPXCount;
        }
        /// <summary>
        /// 獲取一個垂直區域內的黑色像素
        /// </summary>
        /// <param name="startX">開始x</param>
        /// <param name="offset">左偏移像素</param>
        /// <returns></returns>
        private int GetBlackPXCountInX(int startX, int offset, Bitmap img)
        {
            int blackPXCount = 0;
            for (int x = startX; x < startX + offset; x++)
            {
                if (x >= img.Width)
                {
                    continue;
                }
                for (var y = 0; y < img.Height; y++)
                {
                    if (img.GetPixel(x, y).ToArgb() == Color.Black.ToArgb())
                    {
                        blackPXCount++;
                    }
                }
            }
            return blackPXCount;
        }
        /// <summary>
        /// 獲取豎向分割點
        /// </summary>
        /// <param name="img"></param>
        /// <returns>List int[xstart xend]</returns>
        private List<int[]> GetXCutPointList(Bitmap img)
        {
            //分割點  List<int[xstart xend]>
            List<int[]> xCutList = new List<int[]>();
            int startX = -1;//-1表示在尋找開始節點
            for (var x = 0; x < img.Width; x++)
            {
                if (startX == -1)//開始點
                {
                    int blackPXCount = GetBlackPXCountInX(x, 2, img);
                    //若是大於有效像素則是開始節點 ,0-x的矩形區域大於3像素，認爲是字母，防止一些噪點被切割           
                    if (blackPXCount > 5)
                    {
                        startX = x;

                    }
                }
                else//結束點
                {
                    if (x == img.Width - 1)//判斷是否最後一列
                    {
                        xCutList.Add(new int[] { startX, x });
                        break;
                    }
                    else if (x >= startX + MinWidthPerChar)//隔開必定距離才能結束分割
                    {
                        int blackPXCount = GetBlackPXCountInX(x, 2, img);//判斷後面區域黑色像素點的個數
                        //小於等於閥值則是結束節點                       
                        if (blackPXCount < 2)
                        {

                            if (x > startX + MaxWidthPerChar)//儘可能控制不執行
                            {
                                //大於最大字符的寬度應該是兩個字符粘連到一塊了 從中間分開
                                int middleX = startX + (x - startX) / 2;
                                xCutList.Add(new int[] { startX, middleX });
                                xCutList.Add(new int[] { middleX + 1, x });
                            }
                            else
                            {
                                //驗證黑色像素是否太少
                                blackPXCount = GetBlackPXCountInX(startX, x - startX, img);
                                if (blackPXCount <= 10)
                                {
                                    startX = -1;//重置開始點
                                }
                                else
                                {
                                    xCutList.Add(new int[] { startX, x });
                                }
                            }
                            startX = -1;//重置開始點
                        }
                    }
                }
            }
            return xCutList;
        }
        /// <summary>
        /// 截取圖像的矩形區域
        /// </summary>
        /// <param name="source">源圖像對應picturebox1</param>
        /// <param name="rect">矩形區域，如上初始化的rect</param>
        /// <returns>矩形區域的圖像</returns>
        private Image AcquireRectangleImage(Image source, Rectangle rect)
        {
            if (source == null || rect.IsEmpty) return null;
            //Bitmap bmSmall = new Bitmap(rect.Width, rect.Height, System.Drawing.Imaging.PixelFormat.Format32bppRgb);
            Bitmap bmSmall = new Bitmap(rect.Width, rect.Height, source.PixelFormat);

            using (Graphics grSmall = Graphics.FromImage(bmSmall))
            {
                grSmall.DrawImage(source,
                                  new System.Drawing.Rectangle(0, 0, bmSmall.Width, bmSmall.Height),
                                  rect,
                                  GraphicsUnit.Pixel);
                grSmall.Dispose();
            }
            return bmSmall;
        }
        #endregion

        #region 圖片識別
        /// <summary>
        /// 返回兩圖比較的類似度 最大1
        /// </summary>
        /// <param name="compareImg">對比圖</param>
        /// <param name="mainImg">要識別的圖</param>
        /// <returns></returns>
        private double CompareImg(Bitmap compareImg, Bitmap mainImg)
        {
            int img1x = compareImg.Width;
            int img1y = compareImg.Height;
            int img2x = mainImg.Width;
            int img2y = mainImg.Height;
            //最小寬度
            double min_x = img1x > img2x ? img2x : img1x;
            //最小高度
            double min_y = img1y > img2y ? img2y : img1y;

            double score = 0;
            //重疊的黑色像素
            for (var x = 0; x < min_x; x++)
            {
                for (var y = 0; y < min_y; y++)
                {
                    if (compareImg.GetPixel(x, y).ToArgb() == Color.Black.ToArgb()
                        && compareImg.GetPixel(x, y).ToArgb() == mainImg.GetPixel(x, y).ToArgb())
                    {
                        score++;
                    }
                }
            }
            double originalBlackCount = 0;
            //對比圖片的黑色像素
            for (var x = 0; x < img1x; x++)
            {
                for (var y = 0; y < img1y; y++)
                {
                    if (Color.Black.ToArgb() == compareImg.GetPixel(x, y).ToArgb())
                    {
                        originalBlackCount++;
                    }
                }
            }
            return score / originalBlackCount;
        }
        /// <summary>
        /// 用全部的學習的圖片對比當前圖，經過黑色和圖片比率獲取最大類似度的字符圖片，從而識別
        /// </summary>
        /// <param name="imgArr">要識別圖片的數組</param>
        /// <returns>識別後的字符串</returns>
        public string RecognizeCheckCodeImg(Bitmap bitImg)
        {
            Bitmap EZHimg = EZH(bitImg);
            Bitmap[] imgArr = SplitImage(EZHimg);
            string returnString = string.Empty;
            for (int i = 0; i < imgArr.Length; i++)
            {
                if (imgArr[i] == null)
                {
                    continue;
                }
                var img = imgArr[i];
                if (img == null)
                {
                    continue;
                }
                string[] detailPathList = Directory.GetDirectories(samplePath);
                if (detailPathList == null || detailPathList.Length == 0)
                {
                    continue;
                }
                string resultString = string.Empty;
                //config.txt 文件中指定了識別字母的順序
                string configPath = samplePath + "config.txt";
                if (!File.Exists(configPath))
                {
                    Console.WriteLine("config.txt文件不存在,沒法識別");
                    return null;
                }
                string configString = File.ReadAllText(configPath);
                double maxRate = 0;//類似度  最大1
                foreach (char resultChar in configString)
                {
                    string charPath = samplePath + resultChar.ToString();//特徵目錄存儲路徑
                    if (!Directory.Exists(charPath))
                    {
                        continue;
                    }
                    string[] fileNameList = Directory.GetFiles(charPath);
                    if (fileNameList == null || fileNameList.Length == 0)
                    {
                        continue;
                    }


                    foreach (string filename in fileNameList)
                    {
                        Bitmap imgSample = new Bitmap(filename);
                        //過濾寬高相差太大的
                        if (Math.Abs(imgSample.Width - img.Width) >= 2
                            || Math.Abs(imgSample.Height - img.Height) >= 3)
                        {
                            continue;
                        }
                        //當前類似度                       
                        double currentRate = CompareImg(imgSample, img);
                        if (currentRate > maxRate)
                        {
                            maxRate = currentRate;
                            resultString = resultChar.ToString();
                        }
                        imgSample.Dispose();
                    }
                }
                returnString = returnString + resultString;
            }
            return returnString;
        }
        #endregion

    }
}

第2、測試程序代碼

using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Net;
using System.Drawing;
using System.IO;

namespace CheckCodeRecognizeLibTest
{
    class Program
    {
        static void Main(string[] args)
        {
            Console.WriteLine("正在下載驗證碼......");
            //CookieContainer cc = new CookieContainer();
            //byte[] imgByte = HttpWebRequestForBPMS.GetWebResorce("http://hd.cnrds.net/hd/login.do?action=createrandimg",cc);
            //MemoryStream ms1 = new MemoryStream(imgByte);
           // Bitmap bm = (Bitmap)Image.FromStream(ms1); 
            Bitmap img = HttpWebRequestForBPMS.GetWebImage("http://hd.cnrds.net/hd/login.do?action=createrandimg");
            Console.WriteLine("驗證碼下載成功，正在識別.....");
            CheckCodeRecognizeLib.CheckCodeRecognize regImg = new CheckCodeRecognizeLib.CheckCodeRecognize();
            string regResult= regImg.RecognizeCheckCodeImg(img);
            Console.WriteLine("驗證碼識別成功，驗證碼結果爲："+regResult);
            
          
        }       
    }
  
    
}