opencv+tesseract破解教务管理验证码

折腾了一天，弄得差不多了，训练器还要继续调试（折腾）。

效果如图：

)}}

• 程序：

using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Threading.Tasks;
using System.IO;
using Emgu.CV;
using Emgu.CV.CvEnum;
using Emgu.CV.Structure;
using System.Drawing;

namespace jwglcode
{
    /// 

    /// 教务管理验证码识别
    /// 
    class jwglCode : IDisposable
    {
        /// 

        /// 原始验证码图片
        /// 
        public Mat Code;
        /// 

        /// 处理后图片
        /// 
        public Mat CodeReduction;
        /// 

        /// 存储单个字符的数组
        /// 
        public Mat[] CodeSingleArr;
        /// 

        /// 存储旋转后单个字符的数组
        /// 
        public Mat[] CodeSingleRotateArr;
        /// 

        /// 验证码长度
        /// 
        public int CodeLength;
        /// 

        /// 验证码数据（解析结果）
        /// 
        public string CodeStr;

        /// 

        /// 验证码识别
        /// 
        /// 原图像
        /// 验证码长度
        public jwglCode(Mat src,int codeLength = 4)
        {
            Code = src;
            CodeLength = codeLength;
            run();
        }

        /// 

        /// 判断数据是否在上下限内
        /// 
        /// 数据
        /// 目标值
        /// 误差
        /// 
        private bool rangeTest(byte data, byte todata, byte r)
        {
            if (data >= todata - r && data <= todata + r)
                return true;
            else
                return false;
        }

        /// 

        /// 过滤干扰点
        /// （如果一个点和他上下左右都不一样，则将此点颜色改为周围颜色）
        /// 
        /// 图像数据
        /// 
        private Mat filter(Mat src)
        {
            Image src1 = src.ToImage();
            for (int i = 1; i 1; i++)
            {
                for (int j = 1; j 1; j++)
                {
                    //richTextBox1.AppendText(string.Format("[{0},{1}] => {2}\n", i, j, src1.Data[i, j, 0]));
                    if (src1.Data[i, j, 0] != src1.Data[i + 1, j, 0] &&
                        src1.Data[i, j, 0] != src1.Data[i - 1, j, 0] &&
                        src1.Data[i, j, 0] != src1.Data[i, j - 1, 0] &&
                        src1.Data[i, j, 0] != src1.Data[i, j + 1, 0])
                    {

                        src1.Data[i, j, 0] = (byte)((byte)255 - src1.Data[i, j, 0]);
                    }
                }
            }
            return src1.Mat;
        }

        /// 

        /// 错误信息
        /// 
        /// 信息
        private void error(string msg)
        {
            //System.Windows.Forms.MessageBox.Show(msg);
            throw new Exception(msg);
        }

        /// 

        /// 提示消息
        /// 
        /// 信息
        private void info(string msg)
        {
            //System.Windows.Forms.MessageBox.Show(msg);
        }

        /// 

        /// 运行全部计算
        /// 
        private void run ()
        {
            //修图
            CodeReduction = picReduction(Code);
            //分割
            CodeSingleArr = SplitPic(CodeReduction);

            //转正
            if (CodeSingleArr.Length == CodeLength)
            {
                List codeSingleRotateList = new List();
                foreach (var code in CodeSingleArr)
                {
                    codeSingleRotateList.Add(rotate(code));
                }
                CodeSingleRotateArr = codeSingleRotateList.ToArray();
            }
            else
            {
                error(string.Format("字符长度错误！目标值：{0}，搜索值：{1}！", CodeLength, CodeSingleArr.Length));
            }

            //识别
            if(CodeSingleRotateArr.Length == CodeLength)
            {
                string codeStr = "";
                foreach(var code in CodeSingleRotateArr)
                {
                    string filename = saveMatByRandomName(code);
                    codeStr += ocrChar(filename);
                }
                CodeStr = codeStr;
            }
            else
            {
                error(string.Format("字符长度错误！目标值：{0}，搜索值：{1}！", CodeLength, CodeSingleArr.Length));
            }
        }

        /// 

        /// 图像处理
        /// 
        /// 图像
        /// 处理后图像
        private Mat picReduction(Mat src)
        {
            Mat ret = new Mat();
            try
            {
                Image src1 = src.ToImage();
                for (int i = 0; i                 {
                    for (int j = 0; j                     {
                        if (rangeTest(src1.Data[i, j, 0], (byte)150, 20) &&
                                rangeTest(src1.Data[i, j, 1], (byte)10, 10) &&
                                rangeTest(src1.Data[i, j, 2], (byte)10, 10))
                        {
                            src1.Data[i, j, 0] = 0;
                            src1.Data[i, j, 1] = 0;
                            src1.Data[i, j, 2] = 0;
                        }
                        else
                        {
                            src1.Data[i, j, 0] = 255;
                            src1.Data[i, j, 1] = 255;
                            src1.Data[i, j, 2] = 255;
                        }
                    }
                }
                CvInvoke.CvtColor(src1.Mat, ret, ColorConversion.Bgr2Gray);
                return filter(ret);
            }
            catch (Exception ex)
            {
                error(ex.Message);
            }
            return new Mat ();
        }

        /// 

        /// 分割图像
        /// 
        /// 原图像
        /// 分割后数组
        private Mat[] SplitPic(Mat src)
        {
            Image m = src.ToImage();
            Imagebyte> img = new Imagebyte>(m.Cols, m.Rows, new Gray(0));

            for (int j = 0; j //垂直投影
            {
                int d = 0;
                for (int i = 0; i //计算投影
                {
                    if (m.Data[i, j, 0] == 0)
                        d++;
                }
                for (int i = 0; i //绘制投影
                {
                    img.Data[i, j, 0] = 255;
                }
                if (d <2)//绘制分割线
                    for (int i = 0; i                     {
                        m.Data[i, j, 0] = 127;
                    }
            }
            //分割字符组
            int charwidth = 0;
            List mats = new List();
            for (int j = 0; j             {
                if (m.Data[m.Rows - 1, j, 0] != 127)//判最后一行，貌似最后一行都没占用
                {
                    charwidth++;
                }
                else
                {
                    if (charwidth != 0)
                    {
                        int picnums = (int)((double)charwidth / 11 + 0.5);//四舍五入
                        picnums = (picnums == 0) ? 1 : picnums;//i/1/l太细了...
                        int Onewidth= (picnums > 1) ? charwidth / picnums + 1 : charwidth / picnums;

                        for (int pic = 0; pic                         {
                            mats.Add(new Mat(m.Mat, new Rectangle(j - charwidth + pic * onewidth, 0, onewidth, m.Rows - 1)));
                        }
                    }
                    charwidth = 0;
                }
            }
            //切字完成
            return mats.ToArray();
        }

        /// 

        /// 图像大小归一化
        /// 
        /// 图像
        /// 图像大小
        /// 归一化图像
        private Mat SizeNormalLize(Mat src,int sizehw = 30)
        {
            Imagebyte> mm = new Imagebyte>(sizehw, sizehw, new Gray(255));
            Image mI = src.ToImage();

            int h = (sizehw - src.Width) / 2, w = (sizehw - src.Height) / 2;
            for (int i = 0; i             {
                for (int j = 0; j                 {
                    if (mI.Data[i, j, 0] == 0)
                        mm.Data[i + w - 1, j + h - 1, 0] = mI.Data[i, j, 0];
                }
            }

            return mm.Mat;
        }

        /// 

        /// 字符图像转正,大小归一化
        /// 
        /// 单个字符图像
        /// /// 单个字符图像
        /// 转正后图像
        private Mat rotate(Mat src,int r_range = 30)
        {
            Image sourceImage = SizeNormalLize(src).ToImage();

            int r_perfect = 0;
            int hLengthMin = 999;
            for (int r = -r_range; r <= r_range; r++)
            {
                Imagebyte> rotateImagetemp = sourceImage.Rotate(r, new Gray(255));

                int hLength = 0;
                for (int j = 0; j                 {
                    for (int i = 0; i                     {
                        if(rotateImagetemp.Data[i,j,0] == 0)
                        {
                            hLength++;//计算水平投影长度
                            break;
                        }
                    }
                }

                hLengthMin = (hLengthMin //保存最小水平投影长度
                r_perfect = (hLengthMin //保存最佳角度
            }

            Imagebyte> rotateImage = sourceImage.Rotate(r_perfect, new Gray(255));
            return rotateImage.Mat;
        }

        /// 

        /// 字符图像转正,大小归一化
        /// 
        /// 单个字符图像
        /// 转正后图像
        [Obsolete("This methods is obsolete; use methods rotate instead.")]
        private Mat rotate2(Mat m)
        {
            int sizehw = 30;
            Imagebyte> mm = new Imagebyte>(sizehw, sizehw, new Gray(255));
            Image mI = m.ToImage();

            int h = (sizehw - m.Width) / 2, w = (sizehw - m.Height) / 2;
            for (int i = 0; i             {
                for (int j = 0; j                 {
                    if (mI.Data[i, j, 0] == 0)
                        mm.Data[i + w - 1, j + h - 1, 0] = mI.Data[i, j, 0];
                }
            }

            Image sourceImage = mm;

            int smallr_l = 999;//正无穷
            int r_perfect = 0;
            int r_range = 30;
            for (int r = -r_range; r <= r_range; r++)
            {
                Imagebyte> rotateImagetemp = sourceImage.Rotate(r, new Gray(255));

                int minj = 999, maxj = 0;
                for (int j = 0; j                 {
                    for (int i = 0; i                     {
                        if (rotateImagetemp.Data[i, j, 0] == 0)
                        {
                            minj = (minj                             maxj = (maxj > j) ? maxj : j;
                        }
                    }
                }
                smallr_l = (smallr_l <(maxj - minj)) ? smallr_l : (maxj - minj);
                r_perfect = (smallr_l <(maxj - minj)) ? r_perfect : r;
            }

            info(string.Format("r = {0},w = {1}\n", r_perfect, smallr_l));
            Imagebyte> rotateImage = sourceImage.Rotate(r_perfect, new Gray(255));
            return rotateImage.Mat;
        }

        /// 

        /// 使用随机名称保存bmp图像
        /// 
        /// 图像
        /// 名称
        private string saveMatByRandomName(Mat src)
        {
            Random seed = new Random();
            string filename = seed.Next().ToString() + ".bmp";
            while (true)
            {
                if (!File.Exists(filename))//文件存在
                {
                    SaveMat(src, filename);
                    break;
                }
                filename = seed.Next().ToString() + ".bmp";
            }
            return filename;
        }

        /// 

        /// 保存图像
        /// 
        /// 图像
        /// 名称
        private void SaveMat (Mat src,string filename)
        {
            src.Save(filename);
        }

        /// 

        /// 识别图像文件中字符
        /// 
        /// 图像名称
        /// 识别结果
        private string ocrChar(string filename)
        {
            RunTesseract(string.Format("-psm 10 -l jwgl {0} out", filename));
            //这个自己训练...懒得训练的直接用eng吧
            //RunTesseract(string.Format("-psm 10 -l jwgl+eng {0} out", filename));
            //RunTesseract(string.Format("-psm 10 -l eng {0} out", filename));
            string r = "";
            try
            {
                FileStream f = File.OpenRead("out.txt");
                StreamReader fr = new StreamReader(f, Encoding.UTF8);
                r = ((char)fr.Read()).ToString();
                //r = fr.ReadLine();
                fr.Close();
                f.Close();
            }
            catch (Exception ex)
            {
                error(ex.Message);
            }
            return r;
        }

        /// 

        /// 调用Tesseract
        /// 
        /// 参数
        /// 读取到的数据
        private static string RunTesseract(string command)
        {
            System.Diagnostics.Process p = new System.Diagnostics.Process();
            p.StartInfo.FileName = "tesseract.exe";
            p.StartInfo.Arguments = command;
            p.StartInfo.UseShellExecute = false;
            p.StartInfo.RedirectStandardInput = true;
            p.StartInfo.RedirectStandardOutput = true;
            p.StartInfo.RedirectStandardError = true;
            p.StartInfo.CreateNoWindow = true;
            bool b = p.Start();
            string output = "";
            output = p.StandardOutput.ReadToEnd();
            p.WaitForExit(5000);
            p.Close();
            return output;
        }

        #region IDisposable Support
        private bool disposedValue = false; // 要检测冗余调用

        protected virtual void Dispose(bool disposing)
        {
            if (!disposedValue)
            {
                if (disposing)
                {
                    // TODO: 释放托管状态(托管对象)。
                    Code.Dispose();
                    CodeReduction.Dispose();
                    foreach (var a in CodeSingleArr)
                        a.Dispose();
                    foreach(var a in CodeSingleRotateArr)
                        a.Dispose();
                }

                // TODO: 释放未托管的资源(未托管的对象)并在以下内容中替代终结器。
                // TODO: 将大型字段设置为 null。

                disposedValue = true;
            }
        }

        // TODO: 仅当以上 Dispose(bool disposing) 拥有用于释放未托管资源的代码时才替代终结器。
        // ~jwglCode() {
        // // 请勿更改此代码。将清理代码放入以上 Dispose(bool disposing) 中。
        // Dispose(false);
        // }

        // 添加此代码以正确实现可处置模式。
        public void Dispose()
        {
            // 请勿更改此代码。将清理代码放入以上 Dispose(bool disposing) 中。
            Dispose(true);
            // TODO: 如果在以上内容中替代了终结器，则取消注释以下行。
            // GC.SuppressFinalize(this);
        }
        #endregion
    }
}