private TextureClass GetBestMatchForFeature(TextureSample sample, TextureFeatures feature)
{
var distancesList = GetDistancesListForFeature(sample, feature);
var indexesList = new int[TextureClassCount];
var classesList = new List <int>();
var classesFlags = new List <int>();
while (true)
{
int bestClass = GetBestDistance(distancesList, indexesList, classesList, classesFlags);
indexesList[bestClass] += 1;
if (indexesList[bestClass] == knownClasses[bestClass].KnownSamplesNumber(feature))
{
classesFlags.Add(bestClass);
if (classesFlags.Count == knownClasses.Count)
{
return(knownClasses[classesFlags[0]]);
}
}
if (indexesList[bestClass] == RecognitionParameters.NeededNeighborsNumber)
{
return(knownClasses[bestClass]);
}
}
}
private RecognitionResult GetBestMatch(TextureSample sample, Rectangle region)
{
var answer = new RecognitionResult(region, sample);
answer.SetAnswer(TextureFeatures.GLCM, GetBestMatchForFeature(sample, TextureFeatures.GLCM));
answer.SetAnswer(TextureFeatures.LBP, GetBestMatchForFeature(sample, TextureFeatures.LBP));
return(answer);
}
private List <double>[] GetDistancesListForFeature(TextureSample sample, TextureFeatures feature)
{
var distancesList = new List <double> [TextureClassCount];
for (int i = 0; i < TextureClassCount; ++i)
{
distancesList[i] = knownClasses[i].GetSortedDistances(sample, feature);
}
return(distancesList);
}
internal List <double> GetSortedDistances(TextureSample sample, TextureFeatures feature)
{
var list = new List <double>();
var count = KnownSamplesNumber(feature);
for (int i = 0; i < count; ++i)
{
list.Add(GetDistance(sample, i, feature));
}
list.Sort();
return(list);
}
private void SplitImageThread(object parameter)
{
SplitImageThreadParameter input = parameter as SplitImageThreadParameter;
if (input == null)
{
throw new ArgumentException("Неправильный формат потока");
}
int fragmentSize = RecognitionParameters.RecognitionFragmentSize;
var image = input.Image;
var regions = input.ResultRegions;
var samples = input.ResultSamples;
int XCount = image.Width / fragmentSize;
int YCount = image.Height / fragmentSize;
for (int y = input.From; (y < input.To) && (!isRecognizingAborted); ++y)
{
for (int x = 0; (x < XCount) && (!isRecognizingAborted); ++x)
{
int fromX = x * fragmentSize;
int fromY = y * fragmentSize;
int toX = fromX + fragmentSize;
int toY = fromY + fragmentSize;
TextureSample sample = null;
try
{
var imageData = image.GetGrayData(fromX, fromY, toX, toY);
sample = new TextureSample(imageData, false);
}
catch (Exception ex)
{
if (MessageBox.Show("Возникла ошибка. Продолжать?\n" + ex.Message, "Внимание", MessageBoxButtons.YesNo) == DialogResult.No)
{
isRecognizingAborted = true;
}
else
{
continue;
}
}
lock ((regions as ICollection).SyncRoot)
{
regions.Add(new Rectangle(fromX, fromY, fragmentSize, fragmentSize));
samples.Add(sample);
recognitionProgress += 0.9 / (XCount * YCount);
}
}
}
input.ResetEvent.Set();
}
private double GetDistance(TextureSample sample, int featureIndex, TextureFeatures feature)
{
switch (feature)
{
case TextureFeatures.GLCM:
return(sample.GLCM.GetDistance(glcmFeatures[featureIndex]));
case TextureFeatures.LBP:
return(sample.LBP.GetDistance(lbpFeatures[featureIndex]));
default:
return(0);
}
}
public List <RecognitionResult> RecognizeImages(List <Bitmap> images)
{
recognitionProgress = 0;
isRecognizing = true;
var answers = new List <RecognitionResult>();
for (int i = 0; i < images.Count; ++i)
{
var imageToRecognize = new SimpleImage(images[i]);
var imageData = imageToRecognize.GetGrayData(0, 0, imageToRecognize.Width, imageToRecognize.Height);
var sample = new TextureSample(imageData, false);
answers.Add(GetBestMatch(sample, new Rectangle(0, 0, imageData.Width, imageData.Height)));
recognitionProgress += 1.0 * i / images.Count;
}
recognitionProgress = 1;
isRecognizing = false;
return(answers);
}
public RecognitionResult RecognizeImage(Bitmap image, Rectangle region)
{
recognitionProgress = 0;
isRecognizing = true;
var imageToRecognize = new SimpleImage(image);
recognitionProgress = 0.2;
var imageData = imageToRecognize.GetGrayData(region.X, region.Y,
region.X + region.Width, region.Y + region.Height);
var sample = new TextureSample(imageData, false);
recognitionProgress = 0.8;
var answer = GetBestMatch(sample, region);
recognitionProgress = 1;
isRecognizing = false;
return(answer);
}
public RecognitionResult RecognizeImage(Bitmap image, Rectangle region)
{
recognitionProgress = 0;
isRecognizing = true;
var imageToRecognize = new SimpleImage(image);
recognitionProgress = 0.2;
var imageData = imageToRecognize.GetGrayData(region.X, region.Y,
region.X + region.Width, region.Y + region.Height);
var sample = new TextureSample(imageData, false);
recognitionProgress = 0.8;
var answer = GetBestMatch(sample, region);
recognitionProgress = 1;
isRecognizing = false;
return answer;
}
private RecognitionResult GetBestMatch(TextureSample sample, Rectangle region)
{
var answer = new RecognitionResult(region, sample);
answer.SetAnswer(TextureFeatures.GLCM, GetBestMatchForFeature(sample, TextureFeatures.GLCM));
answer.SetAnswer(TextureFeatures.LBP, GetBestMatchForFeature(sample, TextureFeatures.LBP));
return answer;
}
public RecognitionResult(Rectangle region, TextureSample sample)
{
this.region = region;
this.sample = sample;
answers = new Dictionary <TextureFeatures, TextureClass>();
}
public RecognitionResult()
{
region = new Rectangle(0, 0, 0, 0);
sample = null;
answers = new Dictionary <TextureFeatures, TextureClass>();
}
public RecognitionResult(Rectangle region, TextureSample sample)
{
this.region = region;
this.sample = sample;
answers = new Dictionary<TextureFeatures, TextureClass>();
}
public RecognitionResult()
{
region = new Rectangle(0, 0, 0, 0);
sample = null;
answers = new Dictionary<TextureFeatures, TextureClass>();
}
private TextureClass GetBestMatchForFeature(TextureSample sample, TextureFeatures feature)
{
var distancesList = GetDistancesListForFeature(sample, feature);
var indexesList = new int[TextureClassCount];
var classesList = new List<int>();
var classesFlags = new List<int>();
while (true)
{
int bestClass = GetBestDistance(distancesList, indexesList, classesList, classesFlags);
indexesList[bestClass] += 1;
if (indexesList[bestClass] == knownClasses[bestClass].KnownSamplesNumber(feature))
{
classesFlags.Add(bestClass);
if (classesFlags.Count == knownClasses.Count)
{
return knownClasses[classesFlags[0]];
}
}
if (indexesList[bestClass] == RecognitionParameters.NeededNeighborsNumber)
{
return knownClasses[bestClass];
}
}
}
private List<double>[] GetDistancesListForFeature(TextureSample sample, TextureFeatures feature)
{
var distancesList = new List<double>[TextureClassCount];
for (int i = 0; i < TextureClassCount; ++i)
{
distancesList[i] = knownClasses[i].GetSortedDistances(sample, feature);
}
return distancesList;
}
private void SplitImageThread(object parameter)
{
SplitImageThreadParameter input = parameter as SplitImageThreadParameter;
if (input == null)
{
throw new ArgumentException("Неправильный формат потока");
}
int fragmentSize = RecognitionParameters.RecognitionFragmentSize;
var image = input.Image;
var regions = input.ResultRegions;
var samples = input.ResultSamples;
int XCount = image.Width / fragmentSize;
int YCount = image.Height / fragmentSize;
for (int y = input.From; (y < input.To) && (!isRecognizingAborted); ++y)
{
for (int x = 0; (x < XCount) && (!isRecognizingAborted); ++x)
{
int fromX = x * fragmentSize;
int fromY = y * fragmentSize;
int toX = fromX + fragmentSize;
int toY = fromY + fragmentSize;
TextureSample sample = null;
try
{
var imageData = image.GetGrayData(fromX, fromY, toX, toY);
sample = new TextureSample(imageData, false);
}
catch (Exception ex)
{
if (MessageBox.Show("Возникла ошибка. Продолжать?\n" + ex.Message, "Внимание", MessageBoxButtons.YesNo) == DialogResult.No)
{
isRecognizingAborted = true;
}
else
{
continue;
}
}
lock ((regions as ICollection).SyncRoot)
{
regions.Add(new Rectangle(fromX, fromY, fragmentSize, fragmentSize));
samples.Add(sample);
recognitionProgress += 0.9 / (XCount * YCount);
}
}
}
input.ResetEvent.Set();
}
public List<RecognitionResult> RecognizeImages(List<Bitmap> images)
{
recognitionProgress = 0;
isRecognizing = true;
var answers = new List<RecognitionResult>();
for (int i = 0; i < images.Count; ++i)
{
var imageToRecognize = new SimpleImage(images[i]);
var imageData = imageToRecognize.GetGrayData(0, 0, imageToRecognize.Width, imageToRecognize.Height);
var sample = new TextureSample(imageData, false);
answers.Add(GetBestMatch(sample, new Rectangle(0, 0, imageData.Width, imageData.Height)));
recognitionProgress += 1.0 * i / images.Count;
}
recognitionProgress = 1;
isRecognizing = false;
return answers;
}
private double GetDistance(TextureSample sample, int featureIndex, TextureFeatures feature)
{
switch (feature)
{
case TextureFeatures.GLCM:
return sample.GLCM.GetDistance(glcmFeatures[featureIndex]);
case TextureFeatures.LBP:
return sample.LBP.GetDistance(lbpFeatures[featureIndex]);
default:
return 0;
}
}
internal List<double> GetSortedDistances(TextureSample sample, TextureFeatures feature)
{
var list = new List<double>();
var count = KnownSamplesNumber(feature);
for (int i = 0; i < count; ++i)
{
list.Add(GetDistance(sample, i, feature));
}
list.Sort();
return list;
}
