/// <summary>
/// Using superimposition and then compare bool arrays.
/// </summary>
/// <param name="inputImage"></param>
/// <returns></returns>
private Classification RecognizeSuperimposedImageWithBoolArrayMethod(Image inputImage)
{
//If training set isn't loaded yet, load it
if (!loadedTrainingSet)
{
LoadTrainingSet();
}
//Fill up rankTable
foreach (TrainingSetUnit tsu in trainingSet)
{
RankEntry re = new RankEntry();
Image tempInputImage = ProcrustesAnalysis.SaveTranslatedImage(inputImage);
Image tempTsuImage = ProcrustesAnalysis.SaveTranslatedImage(tsu.image);
bool[,] array1 = ImageHelper.MonoBitmapToBoolArray((Bitmap)tempInputImage);
bool[,] array2 = ImageHelper.MonoBitmapToBoolArray((Bitmap)tempTsuImage);
re.squareDistance = SquareDistanceBetweenBoolArrays(array1, array2);
re.classification = tsu.classification;
rankTable.Add(re);
}
rankTable.Sort();
Dictionary<Classification, int> resultCountTable = new Dictionary<Classification, int>();
for (int i = 0; i < k; i++)
{
try
{
resultCountTable.Add(rankTable[i].classification, 0);
}
catch (ArgumentException)
{
resultCountTable[rankTable[i].classification]++;
}
}
Classification MaxLikelihoodClassification = rankTable[0].classification;
int MaxLikelihoodClassificationCount = 1;
foreach (var entry in resultCountTable)
{
if (entry.Value > MaxLikelihoodClassificationCount)
{
MaxLikelihoodClassification = entry.Key;
MaxLikelihoodClassificationCount = entry.Value;
}
}
rankTable.Clear();
return MaxLikelihoodClassification;
}
/// <summary>
/// Applying KNN algorithm to two images, using Hausdorff method as difference.
/// </summary>
/// <param name="inputImage"></param>
/// <returns></returns>
private Classification RecognizeSuperimposedImageWithHausdorffMethod(Image inputImage)
{
//If training set isn't loaded yet, load it
if (!loadedTrainingSet)
{
LoadTrainingSet();
}
//Fill up rankTable
foreach (TrainingSetUnit tsu in trainingSet)
{
RankEntry re = new RankEntry();
re.squareDistance = ProcrustesAnalysis.ComputeHausdroffDistance(inputImage, tsu.image);
re.classification = tsu.classification;
rankTable.Add(re);
}
rankTable.Sort();
Dictionary<Classification, int> resultCountTable = new Dictionary<Classification, int>();
for (int i = 0; i < k; i++)
{
try
{
resultCountTable.Add(rankTable[i].classification, 0);
}
catch (ArgumentException)
{
resultCountTable[rankTable[i].classification]++;
}
}
Classification MaxLikelihoodClassification = rankTable[0].classification;
int MaxLikelihoodClassificationCount = 1;
foreach (var entry in resultCountTable)
{
if (entry.Value > MaxLikelihoodClassificationCount)
{
MaxLikelihoodClassification = entry.Key;
MaxLikelihoodClassificationCount = entry.Value;
}
}
rankTable.Clear();
return MaxLikelihoodClassification;
}
/// <summary>
/// Applying KNN algorithm to two bool arrays.
/// </summary>
/// <param name="boolArray"></param>
/// <returns></returns>
private Classification RecognizeImageWithBoolArrayMethod(Image img)
{
//If training set isn't loaded yet, load it
if (!loadedTrainingSet)
{
LoadTrainingSet();
}
//Fill up rankTable
foreach (TrainingSetUnit tsu in trainingSet)
{
RankEntry re = new RankEntry();
bool[,] boolArray = ImageHelper.MonoBitmapToBoolArray((Bitmap)img);
bool[,] pixelBoolArray = ImageHelper.MonoBitmapToBoolArray((Bitmap)tsu.image);
re.squareDistance = SquareDistanceBetweenBoolArrays(pixelBoolArray, boolArray);
re.classification = tsu.classification;
rankTable.Add(re);
}
rankTable.Sort();
Dictionary<Classification, int> resultCountTable = new Dictionary<Classification, int>();
for (int i = 0; i < k; i++)
{
try
{
resultCountTable.Add(rankTable[i].classification, 0);
}
catch (ArgumentException)
{
resultCountTable[rankTable[i].classification]++;
}
}
Classification MaxLikelihoodClassification = rankTable[0].classification;
int MaxLikelihoodClassificationCount = 1;
foreach (var entry in resultCountTable)
{
if (entry.Value > MaxLikelihoodClassificationCount)
{
MaxLikelihoodClassification = entry.Key;
MaxLikelihoodClassificationCount = entry.Value;
}
}
rankTable.Clear();
return MaxLikelihoodClassification;
}