private void ImportTrainingSet()
{
// Create an instance of the open file dialog box.
var openFileDialog = new OpenFileDialog
{
Filter = @"Txt Files (.txt)|*.txt",
FilterIndex = 1
};
// Call the ShowDialog method to show the dialog box.
var userClickedOk = openFileDialog.ShowDialog();
// Process input if the user clicked OK.
if (userClickedOk == DialogResult.OK)
{
FeaturesViewModel.RemoveAllFeatures();
using (var file = new StreamReader(openFileDialog.FileName))
{
int numLandcoverTypes = int.Parse(file.ReadLine());
Dictionary <int, LandcoverTypeViewModel> landCoverTypes = new Dictionary <int, LandcoverTypeViewModel>();
for (int i = 0; i < numLandcoverTypes; ++i)
{
int id = int.Parse(file.ReadLine());
string name = file.ReadLine();
Color color = (Color)ColorConverter.ConvertFromString(file.ReadLine());
LandcoverTypeViewModel landcoverType = new LandcoverTypeViewModel(id, name, color);
landCoverTypes.Add(id, landcoverType);
}
int numLayers = int.Parse(file.ReadLine());
var layers = new List <CreateLayerViewModel>();
for (int i = 0; i < numLayers; ++i)
{
string path = file.ReadLine();
bool contrastEnhancement = bool.Parse(file.ReadLine());
SatelliteType satelliteType = (SatelliteType)Enum.Parse(typeof(SatelliteType), file.ReadLine());
bool isRed = bool.Parse(file.ReadLine());
bool isGreen = bool.Parse(file.ReadLine());
bool isBlue = bool.Parse(file.ReadLine());
double minCutPercentage = double.Parse(file.ReadLine());
double maxCutPercentage = double.Parse(file.ReadLine());
layers.Add(new CreateLayerViewModel(path, contrastEnhancement, satelliteType, isRed, isGreen, isBlue, minCutPercentage, maxCutPercentage));;
}
var missingLayers = layers.Where(s => _mainWindowViewModel.Layers.All(b => b.Path != s.Path)).ToList();
// Check if some bands have to be added.
if (missingLayers.Count > 0)
{
var dialog = new AddLayersDialog();
if (dialog.ShowImportMissingBandsDialog(missingLayers) == true && dialog.DialogViewModel.Layers.Count > 0)
{
var loadBandsTask = new LoadBandsTask(_mainWindowViewModel, dialog.DialogViewModel);
_mainWindowViewModel.ExecuteLongRunningTask(loadBandsTask);
}
else
{
// Abort import
return;
}
}
_mainWindowViewModel.LandcoverTypes = landCoverTypes.ToImmutableDictionary();
_mainWindowViewModel.SelectedLandCoverTypeViewModel = _mainWindowViewModel.LandcoverTypes.Values.FirstOrDefault();
string featureLine;
while ((featureLine = file.ReadLine()) != null)
{
string[] positionTypeIntensities = featureLine.Split(' ');
string position = positionTypeIntensities[0];
string[] coordinates = position.Split(',');
var landCoverType = _mainWindowViewModel.LandcoverTypes.Values.ToList().FindIndex(t => t.Id == int.Parse(positionTypeIntensities[1]));
var intensities = positionTypeIntensities[2].Split(';').Select(ushort.Parse).ToArray();
FeaturesViewModel.AddFeature(new ClassifiedFeatureVectorViewModel(new ClassifiedFeatureVector(landCoverType, new FeatureVector(intensities), new Point(double.Parse(coordinates[0]), double.Parse(coordinates[1])))));
}
}
}
}
private void ApplyMajorityFilter()
{
NotBlocking = false;
var filtered = new int[_predictionHeight][];
var majorityFilter = Task.Factory.StartNew(() => Parallel.For(0, _predictionHeight, y =>
{
var line = new int[_predictionWidth];
var neighbors = new int[Enum.GetValues(typeof(LandcoverTypeViewModel)).Length - 1];
for (var x = 0; x < _predictionWidth; ++x)
{
if (y > 0 && x > 0 && y < _predictionHeight - 1 && x < _predictionWidth - 1)
{
for (var i = 0; i < neighbors.Length; ++i)
{
neighbors[i] = 0;
}
neighbors[_classification[y - 1][x - 1]]++;
neighbors[_classification[y - 1][x + 0]]++;
neighbors[_classification[y - 1][x + 1]]++;
neighbors[_classification[y + 0][x - 1]]++;
neighbors[_classification[y + 0][x + 0]]++;
neighbors[_classification[y + 0][x + 1]]++;
neighbors[_classification[y + 1][x - 1]]++;
neighbors[_classification[y + 1][x + 0]]++;
neighbors[_classification[y + 1][x + 1]]++;
var biggest = neighbors[0];
var biggestIndex = 0;
for (var index = 0; index < neighbors.Length; index++)
{
if (neighbors[index] > biggest)
{
biggest = neighbors[index];
biggestIndex = index;
}
}
line[x] = biggestIndex;
}
else
{
line[x] = _classification[y][x];
}
}
filtered[y] = line;
}));
majorityFilter.ContinueWith(t =>
{
_classification = filtered;
var stride = _predictionWidth * 4;
var size = _predictionHeight * stride;
var imageData = new byte[size];
var types = SimpleIoc.Default.GetInstance <MainWindowViewModel>().LandcoverTypes;
Parallel.For(0, _predictionWidth, x =>
{
for (var y = 0; y < _predictionHeight; ++y)
{
var index = 4 * y * _predictionWidth + 4 * x;
LandcoverTypeViewModel typeViewModel = types[_classification[y][x]];
var color = typeViewModel.Color;
imageData[index + 0] = color.B;
imageData[index + 1] = color.G;
imageData[index + 2] = color.R;
imageData[index + 3] = color.A;
}
});
Application.Current.Invoke(() =>
{
ClassificationOverlay = BitmapSource.Create(_predictionWidth, _predictionHeight, 96, 96,
PixelFormats.Bgra32, null, imageData, stride);
NotBlocking = true;
ProgressVisibility = Visibility.Hidden;
});
});
}
public FeatureByTypeViewModel(LandcoverTypeViewModel landCoverType)
{
LandCoverType = landCoverType;
Features = new ObservableCollection <ClassifiedFeatureVectorViewModel>();
}
public ObservableCollection <ClassifiedFeatureVectorViewModel> GetFeaturesByType(LandcoverTypeViewModel typeViewModel)
{
return(FeaturesByType.First(f => f.LandCoverType == typeViewModel).Features);
}
public ClassifiedFeatureVectorViewModel(ClassifiedFeatureVector classifiedFeatureVector)
{
ClassifiedFeatureVector = classifiedFeatureVector;
FeatureTypeViewModel = MainWindowViewModel.Default.LandcoverTypes.Values.ToList()[ClassifiedFeatureVector.FeatureClass];
}
