public static SampleCollection LoadPosSamples(string dir, bool isPositive, ColorType colorType)
{
if (false == Directory.Exists(dir))
{
throw new DirectoryNotFoundException(string.Format("目录\"{0}\"不存在", dir));
}
string[] files = ListImageFiles(dir);
SampleCollection coll = new SampleCollection(files.Length * 4);
foreach (string filename in files)
{
Image <Bgr, Byte> img = new Image <Bgr, Byte>(filename);
ISample sample;
sample = new HaarSample(img, isPositive, colorType);
coll.Add(sample);
img._Flip(FLIP.HORIZONTAL);
sample = new HaarSample(img, isPositive, colorType);
coll.Add(sample);
img._Flip(FLIP.VERTICAL);
sample = new HaarSample(img, isPositive, colorType);
coll.Add(sample);
img._Flip(FLIP.HORIZONTAL);
sample = new HaarSample(img, isPositive, colorType);
coll.Add(sample);
}
return(coll);
}
public static SampleCollection LoadNegSamples(string dir, ColorType colorType, Size windowSize)
{
if (false == Directory.Exists(dir))
{
throw new DirectoryNotFoundException(string.Format("目录\"{0}\"不存在", dir));
}
string[] files = ListImageFiles(dir);
SampleCollection coll = new SampleCollection(0);
coll._negIndex = new int[files.Length + 1];
List <HaarSample[]> negList = new List <HaarSample[]>(files.Length);
int i = 0;
foreach (string filename in files)
{
HaarSample[] samples = HaarSample.LoadNegSample(filename, colorType, windowSize, coll._negIndex[i]);
negList.Add(samples);
//if (i + 1 < coll._negIndex.Length)
coll._negIndex[i + 1] = coll._negIndex[i] + samples.Length;
i++;
}
coll.Capacity = coll._negIndex[i];
foreach (HaarSample[] array in negList)
{
coll._samples.AddRange(array);
}
coll._negCount = coll._samples.Count;
return(coll);
}
public Point[] Detect(Image <Bgr, Byte> img, int classifierId)
{
DetectHaarSample sample = new DetectHaarSample(img);
int width = sample.Size.Width;
int height = sample.Size.Height;
Size window = _cascade.Size;
int yEnd = height - window.Height;
int xEnd = width - window.Width;
List <Point> result = new List <Point>((int)Math.Sqrt((xEnd + 1) * (yEnd + 1)));
for (int y = 0; y <= yEnd; y += _yStep)
{
for (int x = 0; x <= xEnd; x += _xStep)
{
Point offset = new Point(x, y);
HaarSample subSample = new HaarSample(sample, offset, window);
if (_cascade.PredictUseOneClassifier(subSample, classifierId))
{
result.Add(offset);
}
}
}
return(result.ToArray());
}
/// <summary>
/// 从一个负样本图像中加载所有样本，共用一个图像
/// </summary>
/// <param name="filename"></param>
/// <param name="colorType"></param>
/// <param name="windowSize">检测窗口大小</param>
/// <returns></returns>
public static HaarSample[] LoadNegSample(string filename, ColorType colorType, Size windowSize, int startIndex)
{
Image <Bgr, Byte> img = new Image <Bgr, Byte>(filename);
MyFloat[,] grayIntergralImage = null, saturationIntergralImage = null, graySquareIntergralImage = null;;
if ((colorType & ColorType.Gray) != 0)
{
Image <Gray, Byte> gray = img.Convert <Gray, Byte>();
Image <Gray, double> grayIntergral, squareIntergral;
//MyFloat[,] norm = NormalizeVariance(gray);
gray.Integral(out grayIntergral, out squareIntergral);
grayIntergralImage = ConvertIntergral(grayIntergral);
graySquareIntergralImage = ConvertIntergral(squareIntergral);
// MyFloat[,] norm = NormalizeVariance(gray);
// grayIntergralImage = CalcIntergarl(norm);
// graySquareIntergralImage= new MyFloat[0,0];
// gray.Integral(out grayIntergral, out squareIntergral);
// grayIntergralImage = ConvertIntergral(grayIntergral);
// graySquareIntergralImage = ConvertIntergral(squareIntergral);
}
if ((colorType & ColorType.Saturation) != 0)
{
Image <Hsv, Byte> hsv = img.Convert <Hsv, Byte>();
Image <Gray, Byte> saturation = hsv[1];
saturationIntergralImage = CalcIntergarl(saturation);
}
int height = img.Rows, width = img.Cols;
int yEnd = height - windowSize.Height + 1, xEnd = width - windowSize.Width + 1;
HaarSample[] samples = new HaarSample[yEnd * xEnd];
int i = 0;
for (int y = 0; y < yEnd; y++)
{
for (int x = 0; x < xEnd; x++)
{
HaarSample s = new HaarSample();
s._grayIntergralImage = grayIntergralImage;
s._saturationIntergralImage = saturationIntergralImage;
s._graySquareIntergralImage = graySquareIntergralImage;
s._isPositive = false;
s._xOffset = x;
s._yOffset = y;
s._id = startIndex + i;
s.CalcMeanAndStd(windowSize);
// s._mean = 0;
// s._std = 1;
samples[i++] = s;
}
}
return(samples);
}