using System;
using System.Collections.Generic;
using System.Diagnostics;
using System.Drawing;
using System.Drawing.Imaging;
using System.IO;
using System.Linq;
using OpenCvSharp;

namespace OCRTest
{
    /// <summary>
    /// OCR训练数据准备工具
    /// 用于准备和预处理训练图片
    /// </summary>
    public class OCRTrainingHelper
    {
        /// <summary>
        /// 批量预处理训练图片（灰度化、二值化、调整大小）
        /// </summary>
        /// <param name="inputFolder">输入文件夹</param>
        /// <param name="outputFolder">输出文件夹</param>
        /// <param name="targetWidth">目标宽度（像素）</param>
        /// <param name="targetHeight">目标高度（像素，0表示自动）</param>
        public static void PreprocessTrainingImages(string inputFolder, string outputFolder, 
            int targetWidth = 800, int targetHeight = 0)
        {
            if (!Directory.Exists(inputFolder))
            {
                throw new DirectoryNotFoundException($"输入文件夹不存在：{inputFolder}");
            }

            if (!Directory.Exists(outputFolder))
            {
                Directory.CreateDirectory(outputFolder);
            }

            var imageFiles = Directory.GetFiles(inputFolder, "*.{png,jpg,jpeg,bmp,tif,tiff}")
                .SelectMany(pattern => Directory.GetFiles(inputFolder, pattern))
                .Distinct()
                .ToArray();

            Console.WriteLine($"找到 {imageFiles.Length} 张图片");

            int successCount = 0;
            int failCount = 0;

            foreach (var inputFile in imageFiles)
            {
                try
                {
                    string fileName = Path.GetFileNameWithoutExtension(inputFile);
                    string outputFile = Path.Combine(outputFolder, $"{fileName}.png");

                    PreprocessSingleImage(inputFile, outputFile, targetWidth, targetHeight);
                    successCount++;

                    Console.WriteLine($"[{successCount + failCount}/{imageFiles.Length}] 处理成功：{fileName}");
                }
                catch (Exception ex)
                {
                    failCount++;
                    Console.WriteLine($"处理失败：{Path.GetFileName(inputFile)} - {ex.Message}");
                }
            }

            Console.WriteLine($"\n处理完成！成功：{successCount}，失败：{failCount}");
        }

        /// <summary>
        /// 预处理单张图片
        /// </summary>
        private static void PreprocessSingleImage(string inputFile, string outputFile, 
            int targetWidth, int targetHeight)
        {
            using (var src = Cv2.ImRead(inputFile, ImreadModes.Color))
            {
                if (src.Empty())
                {
                    throw new Exception("无法读取图片");
                }

                Mat processed = src.Clone();

                // 1. 调整大小
                if (targetWidth > 0 || targetHeight > 0)
                {
                    double scale;
                    if (targetHeight == 0)
                    {
                        scale = (double)targetWidth / src.Width;
                        targetHeight = (int)(src.Height * scale);
                    }
                    else if (targetWidth == 0)
                    {
                        scale = (double)targetHeight / src.Height;
                        targetWidth = (int)(src.Width * scale);
                    }
                    else
                    {
                        scale = Math.Min((double)targetWidth / src.Width, (double)targetHeight / src.Height);
                        targetWidth = (int)(src.Width * scale);
                        targetHeight = (int)(src.Height * scale);
                    }

                    Cv2.Resize(processed, processed, new OpenCvSharp.Size(targetWidth, targetHeight));
                }

                // 2. 转换为灰度图
                if (processed.Channels() > 1)
                {
                    Cv2.CvtColor(processed, processed, ColorConversionCodes.BGR2GRAY);
                }

                // 3. 降噪（高斯模糊）
                Cv2.GaussianBlur(processed, processed, new OpenCvSharp.Size(3, 3), 0);

                // 4. 自适应二值化
                Cv2.AdaptiveThreshold(processed, processed, 255, 
                    AdaptiveThresholdTypes.GaussianC, ThresholdTypes.Binary, 11, 2);

                // 5. 保存为PNG（无损压缩）
                Cv2.ImWrite(outputFile, processed);
            }
        }

        /// <summary>
        /// 生成训练样本的标注文件
        /// </summary>
        /// <param name="imageFolder">图片文件夹</param>
        /// <param name="labels">标签字典（文件名 -> 文本内容）</param>
        public static void GenerateLabelFiles(string imageFolder, Dictionary<string, string> labels)
        {
            foreach (var kvp in labels)
            {
                string imageName = kvp.Key;
                string text = kvp.Value;

                // 支持多种图片格式
                string[] extensions = { ".png", ".jpg", ".jpeg", ".bmp", ".tif", ".tiff" };
                string imagePath = null;

                foreach (var ext in extensions)
                {
                    string path = Path.Combine(imageFolder, imageName + ext);
                    if (File.Exists(path))
                    {
                        imagePath = path;
                        break;
                    }
                }

                if (imagePath == null)
                {
                    Console.WriteLine($"警告：找不到图片文件 {imageName}");
                    continue;
                }

                // 生成对应的txt文件
                string txtPath = Path.ChangeExtension(imagePath, ".txt");
                File.WriteAllText(txtPath, text, System.Text.Encoding.UTF8);

                Console.WriteLine($"生成标注文件：{Path.GetFileName(txtPath)}");
            }
        }

        /// <summary>
        /// 验证训练数据集的完整性
        /// </summary>
        /// <param name="folder">数据集文件夹</param>
        /// <returns>缺失标注文件的图片列表</returns>
        public static List<string> ValidateDataset(string folder)
        {
            var missingLabels = new List<string>();

            var imageFiles = Directory.GetFiles(folder, "*.{png,jpg,jpeg,bmp,tif,tiff}")
                .SelectMany(pattern => Directory.GetFiles(folder, pattern))
                .Distinct()
                .ToArray();

            foreach (var imageFile in imageFiles)
            {
                string txtFile = Path.ChangeExtension(imageFile, ".txt");
                
                if (!File.Exists(txtFile))
                {
                    missingLabels.Add(Path.GetFileName(imageFile));
                }
            }

            if (missingLabels.Count > 0)
            {
                Console.WriteLine($"发现 {missingLabels.Count} 个缺少标注文件的图片：");
                foreach (var file in missingLabels)
                {
                    Console.WriteLine($"  - {file}");
                }
            }
            else
            {
                Console.WriteLine($"数据集完整！共 {imageFiles.Length} 个样本");
            }

            return missingLabels;
        }

        /// <summary>
        /// 统计字符频率（用于生成字符集）
        /// </summary>
        /// <param name="labelFolder">标注文件夹</param>
        /// <returns>字符频率字典</returns>
        public static Dictionary<char, int> AnalyzeCharacterFrequency(string labelFolder)
        {
            var charFrequency = new Dictionary<char, int>();

            var txtFiles = Directory.GetFiles(labelFolder, "*.txt");

            foreach (var txtFile in txtFiles)
            {
                string content = File.ReadAllText(txtFile, System.Text.Encoding.UTF8);

                foreach (char c in content)
                {
                    if (charFrequency.ContainsKey(c))
                    {
                        charFrequency[c]++;
                    }
                    else
                    {
                        charFrequency[c] = 1;
                    }
                }
            }

            // 按频率排序
            var sorted = charFrequency.OrderByDescending(kvp => kvp.Value).ToDictionary(kvp => kvp.Key, kvp => kvp.Value);

            Console.WriteLine("字符频率统计（前50个）：");
            int count = 0;
            foreach (var kvp in sorted.Take(50))
            {
                string displayChar = kvp.Key == ' ' ? "(空格)" : kvp.Key.ToString();
                Console.WriteLine($"  '{displayChar}': {kvp.Value} 次");
                count++;
            }

            Console.WriteLine($"\n总字符种类：{sorted.Count}");

            return sorted;
        }

        /// <summary>
        /// 生成训练脚本（Windows批处理）
        /// </summary>
        /// <param name="fontName">字体名称</param>
        /// <param name="outputPath">输出路径</param>
        public static void GenerateTrainingScript(string fontName, string outputPath)
        {
            string script = $@"@echo off
echo ========================================
echo Tesseract OCR 训练脚本
echo 字体名称：{fontName}
echo ========================================
echo.

set LANG={fontName}
set TESSDATA_PREFIX=%~dp0tessdata

echo [1/8] 生成box文件...
tesseract %LANG%.exp0.tif %LANG%.exp0 batch.nochop makebox
if errorlevel 1 goto error

echo [2/8] 生成训练文件...
tesseract %LANG%.exp0.tif %LANG%.exp0 nobatch box.train
if errorlevel 1 goto error

echo [3/8] 提取字符集...
unicharset_extractor %LANG%.exp0.box
if errorlevel 1 goto error

echo [4/8] 形状聚类...
shapeclustering -F unicharset -O unicharset %LANG%.exp0.tr
if errorlevel 1 goto error

echo [5/8] MF训练...
mftraining -F unicharset -U unicharset -O {fontName}.unicharset %LANG%.exp0.tr
if errorlevel 1 goto error

echo [6/8] CN训练...
cntraining %LANG%.exp0.tr
if errorlevel 1 goto error

echo [7/8] 重命名文件...
rename normproto {fontName}.normproto
rename inttemp {fontName}.inttemp
rename pffmtable {fontName}.pffmtable
rename shapetable {fontName}.shapetable

echo [8/8] 合并训练数据...
combine_tessdata {fontName}.
if errorlevel 1 goto error

echo.
echo ========================================
echo 训练完成！
echo 生成的文件：{fontName}.traineddata
echo 请将其复制到 tessdata 文件夹
echo ========================================
pause
exit /b 0

:error
echo.
echo 训练失败！错误代码：%errorlevel%
pause
exit /b 1
";

            File.WriteAllText(outputPath, script, System.Text.Encoding.UTF8);
            Console.WriteLine($"训练脚本已生成：{outputPath}");
        }

        /// <summary>
        /// 创建示例训练项目
        /// </summary>
        /// <param name="projectFolder">项目文件夹</param>
        public static void CreateSampleProject(string projectFolder)
        {
            if (!Directory.Exists(projectFolder))
            {
                Directory.CreateDirectory(projectFolder);
            }

            // 创建子文件夹
            string rawFolder = Path.Combine(projectFolder, "raw_images");
            string processedFolder = Path.Combine(projectFolder, "processed_images");
            string labelsFolder = Path.Combine(projectFolder, "labels");

            Directory.CreateDirectory(rawFolder);
            Directory.CreateDirectory(processedFolder);
            Directory.CreateDirectory(labelsFolder);

            // 创建README
            string readme = $@"# OCR训练项目

## 文件夹说明
- **raw_images/**: 原始训练图片
- **processed_images/**: 预处理后的图片
- **labels/**: 标注文件（.txt格式）

## 使用步骤

### 1. 准备训练数据
将训练图片放入 `raw_images` 文件夹

### 2. 预处理图片
```csharp
OCRTrainingHelper.PreprocessTrainingImages(
    ""{rawFolder}"",
    ""{processedFolder}"",
    targetWidth: 800
);
```

### 3. 创建标注文件
为每张图片创建对应的 .txt 文件，内容为识别文本

### 4. 验证数据集
```csharp
OCRTrainingHelper.ValidateDataset(""{processedFolder}"");
```

### 5. 分析字符频率
```csharp
OCRTrainingHelper.AnalyzeCharacterFrequency(""{labelsFolder}"");
```

### 6. 生成训练脚本
```csharp
OCRTrainingHelper.GenerateTrainingScript(
    ""myfont"",
    ""{projectFolder}/train.bat""
);
```

### 7. 执行训练
运行 `train.bat` 开始训练

## 注意事项
- 每个样本至少需要100-500张图片
- 图片应覆盖所有需要识别的字符
- 标注文件必须与图片同名（仅扩展名不同）
- 建议使用PNG格式保存预处理后的图片
";

            File.WriteAllText(Path.Combine(projectFolder, "README.md"), readme);

            // 生成训练脚本
            GenerateTrainingScript("myfont", Path.Combine(projectFolder, "train.bat"));

            Console.WriteLine($"示例项目已创建：{projectFolder}");
            Console.WriteLine("请按照 README.md 中的说明进行操作");
        }
    }
}