public void TestBOWKmeansTrainer()
{
Image <Gray, byte> box = EmguAssert.LoadImage <Gray, byte>("box.png");
AKAZE detector = new AKAZE();
VectorOfKeyPoint kpts = new VectorOfKeyPoint();
Mat descriptors = new Mat();
detector.DetectAndCompute(box, null, kpts, descriptors, false);
BOWKMeansTrainer trainer = new BOWKMeansTrainer(100, new MCvTermCriteria(), 3, CvEnum.KMeansInitType.PPCenters);
trainer.Add(descriptors);
Mat vocabulary = new Mat();
trainer.Cluster(vocabulary);
BFMatcher matcher = new BFMatcher(DistanceType.L2);
BOWImgDescriptorExtractor extractor = new BOWImgDescriptorExtractor(detector, matcher);
extractor.SetVocabulary(vocabulary);
Mat descriptors2 = new Mat();
extractor.Compute(box, kpts, descriptors2);
}
// Loadイベント（Formの立ち上げ時に実行）
private void Form1_Load(object sender, EventArgs e)
{
pictureBoxResult.Visible = false;
mat = Cv2.ImRead(@"C:\Users\massi\Downloads\one.jpeg");//比較元画像
//特徴量の検出と特徴量ベクトルの計算
akaze.DetectAndCompute(mat, null, out key_point1, descriptor1);
//画像１の特徴点をoutput1に出力
Cv2.DrawKeypoints(mat, key_point1, output1);
//Cv2.ImShow("output1", output1);
Debug.WriteLine("Formのロード開始");
this.getCameraInfo();
}
private Mat Process(ref Mat buffer)
{
Mat img = new Mat();
AKAZE akaze = AKAZE.Create();
akaze.Threshold = 0.0001;
KeyPoint[] keyPoints;
DMatch[] matches;
List <DMatch> goodMatches = new List <DMatch>();
Mat descriptor = new Mat();
DescriptorMatcher matcher = DescriptorMatcher.Create("BruteForce");
Cv2.CvtColor(buffer, buffer, ColorConversionCodes.BGR2GRAY);
akaze.DetectAndCompute(buffer, null, out keyPoints, descriptor);
Cv2.DrawKeypoints(buffer, keyPoints, img, Scalar.Black);
Cv2.ImShow("keyps", img);
if (islastSeted)
{
matches = matcher.Match(descriptor, lastDescriptor);
for (int i = 0; i < matches.Length; i++)
{
if (matches[i].Distance < distanceStandard)
{
goodMatches.Add(matches[i]);
}
}
//Cv2.DrawMatches(buffer, keyPoints, lastBuffer, lastkeyPoints, goodMatches, img);
img = buffer;
if (goodMatches.Count > 3)
{
float[] average = new float[2];
average[0] = 0; average[1] = 0;
for (int i = 0; i < goodMatches.Count; i++)
{
average[0] += keyPoints[goodMatches[0].QueryIdx].Pt.X -
lastkeyPoints[goodMatches[0].TrainIdx].Pt.X;
average[1] += keyPoints[goodMatches[0].QueryIdx].Pt.Y -
lastkeyPoints[goodMatches[0].TrainIdx].Pt.Y;
}
lastPoint = new Point(lastPoint.X + average[0] / goodMatches.Count, lastPoint.Y + average[1] / goodMatches.Count);
lastBuffer = buffer;
lastDescriptor = descriptor;
lastkeyPoints = keyPoints;
}
Cv2.Circle(img, lastPoint, 15, Scalar.Red, 3);
}
else
{
islastSeted = true;
img = buffer;
lastPoint = new Point(buffer.Cols / 2, buffer.Rows / 2);
lastBuffer = buffer;
lastDescriptor = descriptor;
lastkeyPoints = keyPoints;
}
return(img);
}
public void TestAkazeBlankImage()
{
AKAZE detector = new AKAZE();
Image <Gray, Byte> img = new Image <Gray, byte>(1024, 900);
VectorOfKeyPoint vp = new VectorOfKeyPoint();
Mat descriptors = new Mat();
detector.DetectAndCompute(img, null, vp, descriptors, false);
}
public void FindContours(string sLeftPictureFile, string sRightPictureFile)
{
Mat tokuLeft = new Mat();
Mat tokuRight = new Mat();
Mat output = new Mat();
AKAZE akaze = AKAZE.Create();
KeyPoint[] keyPointsLeft;
KeyPoint[] keyPointsRight;
Mat descriptorLeft = new Mat();
Mat descriptorRight = new Mat();
DescriptorMatcher matcher; //マッチング方法
DMatch[] matches; //特徴量ベクトル同士のマッチング結果を格納する配列
//画像をグレースケールとして読み込み、平滑化する
Mat Lsrc = new Mat(sLeftPictureFile, ImreadModes.Color);
//画像をグレースケールとして読み込み、平滑化する
Mat Rsrc = new Mat(sRightPictureFile, ImreadModes.Color);
//特徴量の検出と特徴量ベクトルの計算
akaze.DetectAndCompute(Lsrc, null, out keyPointsLeft, descriptorLeft);
akaze.DetectAndCompute(Rsrc, null, out keyPointsRight, descriptorRight);
//画像1の特徴点をoutput1に出力
Cv2.DrawKeypoints(Lsrc, keyPointsLeft, tokuLeft);
Image imageLeftToku = BitmapConverter.ToBitmap(tokuLeft);
pictureBox3.SizeMode = PictureBoxSizeMode.Zoom;
pictureBox3.Image = imageLeftToku;
tokuLeft.SaveImage("result/LeftToku.jpg");
//画像2の特徴点をoutput1に出力
Cv2.DrawKeypoints(Rsrc, keyPointsRight, tokuRight);
Image imageRightToku = BitmapConverter.ToBitmap(tokuRight);
pictureBox4.SizeMode = PictureBoxSizeMode.Zoom;
pictureBox4.Image = imageRightToku;
tokuRight.SaveImage("result/RightToku.jpg");
//総当たりマッチング
matcher = DescriptorMatcher.Create("BruteForce");
matches = matcher.Match(descriptorLeft, descriptorRight);
Cv2.DrawMatches(Lsrc, keyPointsLeft, Rsrc, keyPointsRight, matches, output);
output.SaveImage(@"result\output.jpg");
int size = matches.Count();
var getPtsSrc = new Vec2f[size];
var getPtsTarget = new Vec2f[size];
int count = 0;
foreach (var item in matches)
{
var ptSrc = keyPointsLeft[item.QueryIdx].Pt;
var ptTarget = keyPointsRight[item.TrainIdx].Pt;
getPtsSrc[count][0] = ptSrc.X;
getPtsSrc[count][1] = ptSrc.Y;
getPtsTarget[count][0] = ptTarget.X;
getPtsTarget[count][1] = ptTarget.Y;
count++;
}
// SrcをTargetにあわせこむ変換行列homを取得する。ロバスト推定法はRANZAC。
var hom = Cv2.FindHomography(
InputArray.Create(getPtsSrc),
InputArray.Create(getPtsTarget),
HomographyMethods.Ransac);
// 行列homを用いてSrcに射影変換を適用する。
Mat WarpedSrcMat = new Mat();
Cv2.WarpPerspective(
Lsrc, WarpedSrcMat, hom,
new OpenCvSharp.Size(Rsrc.Width, Rsrc.Height));
WarpedSrcMat.SaveImage(@"result\Warap.jpg");
//画像1の特徴点をoutput1に出力
Image imageLeftSyaei = BitmapConverter.ToBitmap(WarpedSrcMat);
pictureBox5.SizeMode = PictureBoxSizeMode.Zoom;
pictureBox5.Image = imageLeftSyaei;
//画像2の特徴点をoutput1に出力
Image imageRightSyaei = BitmapConverter.ToBitmap(Rsrc);
pictureBox6.SizeMode = PictureBoxSizeMode.Zoom;
pictureBox6.Image = imageRightSyaei;
Mat LmatFloat = new Mat();
WarpedSrcMat.ConvertTo(LmatFloat, MatType.CV_16SC3);
Mat[] LmatPlanes = LmatFloat.Split();
Mat RmatFloat = new Mat();
Rsrc.ConvertTo(RmatFloat, MatType.CV_16SC3);
Mat[] RmatPlanes = RmatFloat.Split();
Mat diff0 = new Mat();
Mat diff1 = new Mat();
Mat diff2 = new Mat();
Cv2.Absdiff(LmatPlanes[0], RmatPlanes[0], diff0);
Cv2.Absdiff(LmatPlanes[1], RmatPlanes[1], diff1);
Cv2.Absdiff(LmatPlanes[2], RmatPlanes[2], diff2);
Cv2.MedianBlur(diff0, diff0, 5);
Cv2.MedianBlur(diff1, diff1, 5);
Cv2.MedianBlur(diff2, diff2, 5);
diff0.SaveImage("result/diff0.jpg");
diff1.SaveImage("result/diff1.jpg");
diff2.SaveImage("result/diff2.jpg");
Mat wiseMat = new Mat();
Cv2.BitwiseOr(diff0, diff1, wiseMat);
Cv2.BitwiseOr(wiseMat, diff2, wiseMat);
wiseMat.SaveImage("result/wiseMat.jpg");
Mat openingMat = new Mat();
Cv2.MorphologyEx(wiseMat, openingMat, MorphTypes.Open, new Mat());
Mat dilationMat = new Mat();
Cv2.Dilate(openingMat, dilationMat, new Mat());
Cv2.Threshold(dilationMat, dilationMat, 100, 255, ThresholdTypes.Binary);
dilationMat.SaveImage(@"result\dilationMat.jpg");
Mat LaddMat = new Mat();
Mat RaddMat = new Mat();
Console.WriteLine(dilationMat.GetType());
Console.WriteLine(Rsrc.GetType());
// dilationMatはグレースケールなので合成先のMatと同じ色空間に変換する
Mat dilationScaleMat = new Mat();
Mat dilationColorMat = new Mat();
Cv2.ConvertScaleAbs(dilationMat, dilationScaleMat);
Cv2.CvtColor(dilationScaleMat, dilationColorMat, ColorConversionCodes.GRAY2RGB);
Cv2.AddWeighted(WarpedSrcMat, 0.3, dilationColorMat, 0.7, 0, LaddMat);
Cv2.AddWeighted(Rsrc, 0.3, dilationColorMat, 0.7, 0, RaddMat);
Image LaddImage = BitmapConverter.ToBitmap(LaddMat);
pictureBox7.SizeMode = PictureBoxSizeMode.Zoom;
pictureBox7.Image = LaddImage;
Image RaddImage = BitmapConverter.ToBitmap(RaddMat);
pictureBox8.SizeMode = PictureBoxSizeMode.Zoom;
pictureBox8.Image = RaddImage;
RaddMat.SaveImage(@"result\Result.jpg");
MessageBox.Show("Done!");
}