//自动样品分割
public static List <CharInfo> SplitePlateForAutoSample(Mat plateMat)
{
List <CharInfo> result = new List <CharInfo>();
List <CharInfo> charInfos_Original_Blue = SplitePlateByOriginal(plateMat, plateMat, PlateColor.蓝牌);
List <CharInfo> charInfos_IndexTransform_Blue = SplitePlateByIndexTransform(plateMat,
PlateColor.蓝牌);
List <CharInfo> charInfos_GammaTransform_Blue = SplitePlateByGammaTransform(plateMat,
PlateColor.蓝牌);
List <CharInfo> charInfos_LogTransform_Blue = SplitePlateByLogTransform(plateMat, PlateColor.蓝牌);
List <CharInfo> charInfos_Blue = new List <CharInfo>();
charInfos_Blue.AddRange(charInfos_Original_Blue.ToArray());
charInfos_Blue.AddRange(charInfos_IndexTransform_Blue.ToArray());
charInfos_Blue.AddRange(charInfos_GammaTransform_Blue.ToArray());
charInfos_Blue.AddRange(charInfos_LogTransform_Blue.ToArray());
int isCharCount = 0;
for (int index = 0; index < charInfos_Blue.Count; index++)
{
CharInfo charInfo = charInfos_Blue[index];
charInfo.PlateChar = PlateChar_SVM.Test(charInfo.OriginalMat);
if (charInfo.PlateChar != PlateChar.非字符)
{
isCharCount++;
}
}
if (isCharCount >= 15)
{
return(charInfos_Blue);
}
//如果用蓝色切分字符，少于15个，就再用黄色尝试切分
List <CharInfo> charInfos_Original_Yellow = SplitePlateByOriginal(plateMat, plateMat, PlateColor.黄牌);
List <CharInfo> charInfos_IndexTransform_Yellow = SplitePlateByIndexTransform(plateMat,
PlateColor.黄牌);
List <CharInfo> charInfos_GammaTransform_Yellow = SplitePlateByGammaTransform(plateMat,
PlateColor.黄牌);
List <CharInfo> charInfos_LogTransform_Yellow = SplitePlateByLogTransform(plateMat,
PlateColor.黄牌);
List <CharInfo> charInfos_Yellow = new List <CharInfo>();
charInfos_Yellow.AddRange(charInfos_Original_Yellow.ToArray());
charInfos_Yellow.AddRange(charInfos_IndexTransform_Yellow.ToArray());
charInfos_Yellow.AddRange(charInfos_GammaTransform_Yellow.ToArray());
charInfos_Yellow.AddRange(charInfos_LogTransform_Yellow.ToArray());
isCharCount = 0;
for (int index = 0; index < charInfos_Yellow.Count; index++)
{
CharInfo charInfo = charInfos_Yellow[index];
charInfo.PlateChar = PlateChar_SVM.Test(charInfo.OriginalMat);
if (charInfo.PlateChar != PlateChar.非字符)
{
isCharCount++;
}
}
if (isCharCount >= 15)
{
return(charInfos_Yellow);
}
return(new List <CharInfo>()); //返回?长度为零的集合
}
public static Mat GetCharSvmHOGFeatures(Mat img)
{
float[] descriptor = PlateChar_SVM.ComputeHogDescriptors(img);
Mat feature = Float2Mat(descriptor);
return(feature);
}
//
public static PlateInfo GetPlateInfo(PlateInfo plateInfo, PlateColor plateColor, CharSplitMethod
splitMethod)
{
PlateInfo result = new PlateInfo();
result.PlateCategory = plateInfo.PlateCategory;
result.OriginalMat = plateInfo.OriginalMat;
result.OriginalRect = plateInfo.OriginalRect;
result.PlateLocateMethod = plateInfo.PlateLocateMethod;
result.PlateColor = plateColor;
List <CharInfo> charInfos = new List <CharInfo>();
switch (splitMethod)
{
case CharSplitMethod.伽马:
charInfos = CharSegment_V3.SplitePlateByGammaTransform(plateInfo.OriginalMat,
plateColor);
break;
case CharSplitMethod.指数:
charInfos = CharSegment_V3.SplitePlateByIndexTransform(plateInfo.OriginalMat, plateColor);
break;
case CharSplitMethod.对数:
charInfos = CharSegment_V3.SplitePlateByLogTransform(plateInfo.OriginalMat, plateColor);
break;
case CharSplitMethod.原图:
default:
charInfos = CharSegment_V3.SplitePlateByOriginal(plateInfo.OriginalMat,
plateInfo.OriginalMat, plateColor);
break;
}
//启动SVM识别字符并将其中的非字符去除
for (int index = charInfos.Count - 1; index >= 0; index--)
{
CharInfo charInfo = charInfos[index];
PlateChar plateChar = PlateChar_SVM.Test(charInfo.OriginalMat);
if (plateChar == PlateChar.非字符)
{
charInfos.RemoveAt(index);
}
charInfo.PlateChar = plateChar;
}
result.CharInfos = charInfos;
//检测
CheckLeftAndRightToRemove(result);
CheckPlateColor(result);
return(result);
}
//对分割好的字符进行识别，并得到字符数组
public static List <CharInfo> SpliteCharsInPlateMat(Mat plateMat, List <Rect> rects)
{
if (PlateChar_SVM.IsReady == false)
{
throw new Exception("字符识别库没有准备好");
}
List <CharInfo> result = new List <CharInfo>();
for (int index = 0; index < rects.Count; index++)
{
Rect rect = rects[index];
rect = Utilities.GetSafeRect(rect, plateMat);
CharInfo charInfo = new CharInfo();
Mat originalMat = plateMat.SubMat(rect);
charInfo.OriginalMat = originalMat;
charInfo.OriginalRect = rect;
charInfo.PlateChar = PlateChar_SVM.Test(originalMat);
result.Add(charInfo);
}
result.Sort(new CharInfoLeftComparer());
return(result);
}
//对字符进行训练
public static void TrainSVMDataForCharRecog(string path)
{
Console.WriteLine("preparing for training data!!!!");
List <TrainStruct> svmData = new List <TrainStruct>();
List <List <string> > ImgFiles = new List <List <string> >();
//for(int index=0;index<10;index++)
//{
// string label = "_" + index.ToString();
// string filePath = @"C:\Users\faiz\Desktop\AI\车牌-字符样本\车牌-字符样本\chars\" + label;
// Console.WriteLine("{0}", filePath);
// List<string> files = getSampleFiles(filePath);
// ImgFiles.Add(files);
//}
for (int index = (int)PlateChar.A; index <= (int)PlateChar._9; index++)
{
if (index >= 28 && index <= 37)
{
string labelForNumber = "_" + (index - 28).ToString();
string filePath = path + labelForNumber;//@"C:\Users\faiz\Desktop\AI\车牌-字符样本\车牌-字符样本\chars\"
Console.WriteLine("{0}", filePath);
List <string> files = getSampleFiles(filePath);
ImgFiles.Add(files);
}
else
{
PlateChar plateChar = (PlateChar)index;
string labelForWord = plateChar.ToString();
string filePath = path + labelForWord;//@"C:\Users\faiz\Desktop\AI\车牌-字符样本\车牌-字符样本\chars\"
Console.WriteLine("{0}", filePath);
List <string> files = getSampleFiles(filePath);
ImgFiles.Add(files);
}
}
//现在主要是要识别广东车牌
List <string> exFiles = getSampleFiles(path + PlateChar.粤.ToString());
ImgFiles.Add(exFiles);
for (int index1 = 0; index1 < ImgFiles.Count - 1; index1++)
{
for (int index2 = 0; index2 < ImgFiles.ElementAt(index1).Count; index2++)
{
TrainStruct trainData;
trainData.file = ImgFiles.ElementAt(index1).ElementAt(index2);
trainData.label = index1 + 2;
svmData.Add(trainData);
}
}
//最后再添加东的数据
for (int index = 0; index < ImgFiles.ElementAt(ImgFiles.Count - 1).Count; index++)
{
TrainStruct trainData;
trainData.file = ImgFiles.ElementAt(ImgFiles.Count - 1).ElementAt(index);
trainData.label = (int)PlateChar.粤;
svmData.Add(trainData);
}
Mat samples = new Mat();
Mat responses = new Mat();
//读取数据并进行处理
for (int index = 0; index < svmData.Count; index++)
{
//以灰度的方式读取图片
Mat img = Cv2.ImRead(svmData[index].file, ImreadModes.Grayscale);
//剔除无用数据
if (img.Data == null)
{
Console.WriteLine("failed to load image {0}", svmData[index].file);
}
//对图片进行二值化
Mat dst = new Mat();
Cv2.Threshold(img, dst, 1, 255, ThresholdTypes.Otsu | ThresholdTypes.Binary);
Mat feature = GetCharSvmHOGFeatures(dst);
//获取HOG特征
feature = feature.Reshape(1, 1);
samples.PushBack(feature);
responses.PushBack(Int2Mat(svmData[index].label));
}
// 训练数据的格式，OpenCV规定 samples 中的数据都是需要32位浮点型
// 因为TrainData::create 第一个参数是规定死的要cv_32F
samples.ConvertTo(samples, MatType.CV_32F);
// samples 将图片和样本标签合并成为一个训练集数据
// 第二个参数的原因是，我们的samples 中的每个图片数据的排列都是一行
if (PlateChar_SVM.Train(samples, responses))
{
Console.WriteLine("Trained success!!!");
PlateChar_SVM.Save(@"E:\工作文件夹（workplace）\VSworkplace\PlateRecog\charRecog.xml");
Console.WriteLine("\".xml\"has been written!!1");
}
else
{
Console.WriteLine("failed to train data");
}
}
