Emgucv SURF特征点的寻找和匹配

使用SURF类和其相关的函数需要添加Emgu.CV.Features2D和Emgu.CV.XFeatures2D这两个命名空间，同时也可以使用cuda模块来进行加速，由于使用Cuda需要安装，但Cuda目前支持的显卡驱动并不兼容win10，所以这里我就简单的说一下通用的方法吧。
这个代码里使用了UMat，UMat应该算是3.0版本中的新特性，根据官方的说法，UMat等价于Opencv中的Mat。具体的步骤就是先计算特征点然后再进行匹配。

public partial class Form1 : Form
    {
        public Form1()
        {
            InitializeComponent();
            Image<Bgra, byte> a = new Image<Bgra, byte>("IMG_20150829_192403.JPG").Resize(0.4,Inter.Area);  //模板
            Image<Bgra, byte> b = new Image<Bgra, byte>("DSC_0437.JPG").Resize(0.4, Inter.Area);  //待匹配的图像

            Mat homography = null;
            Mat mask = null;
            VectorOfKeyPoint modelKeyPoints = new VectorOfKeyPoint();
            VectorOfKeyPoint observedKeyPoints = new VectorOfKeyPoint();
            VectorOfVectorOfDMatch matches = new VectorOfVectorOfDMatch();

            UMat a1 = a.Mat.ToUMat(AccessType.Read);
            UMat b1 = b.Mat.ToUMat(AccessType.Read);

            SURF surf = new SURF(300);
            UMat modelDescriptors = new UMat();
            UMat observedDescriptors = new UMat();

            surf.DetectAndCompute(a1, null, modelKeyPoints, modelDescriptors, false);       //进行检测和计算，把opencv中的两部分和到一起了，分开用也可以
            surf.DetectAndCompute(b1, null, observedKeyPoints, observedDescriptors, false);

            BFMatcher matcher = new BFMatcher(DistanceType.L2);       //开始进行匹配
            matcher.Add(modelDescriptors);
            matcher.KnnMatch(observedDescriptors, matches, 2, null);
            mask = new Mat(matches.Size, 1, DepthType.Cv8U, 1);
            mask.SetTo(new MCvScalar(255));
            Features2DToolbox.VoteForUniqueness(matches, 0.8, mask);   //去除重复的匹配

            int Count = CvInvoke.CountNonZero(mask);      //用于寻找模板在图中的位置
            if (Count >= 4)
            {
                Count = Features2DToolbox.VoteForSizeAndOrientation(modelKeyPoints, observedKeyPoints, matches, mask, 1.5, 20);
                if (Count >= 4)
                    homography = Features2DToolbox.GetHomographyMatrixFromMatchedFeatures(modelKeyPoints, observedKeyPoints, matches, mask, 2);
            }

            Mat result = new Mat();
            Features2DToolbox.DrawMatches(a.Convert<Gray, byte>().Mat, modelKeyPoints, b.Convert<Gray, byte>().Mat, observedKeyPoints,matches, result, new MCvScalar(255, 0, 255), new MCvScalar(0, 255, 255), mask);
            //绘制匹配的关系图
            if (homography != null)     //如果在图中找到了模板，就把它画出来
            {
                Rectangle rect = new Rectangle(Point.Empty, a.Size);
                PointF[] points = new PointF[]
                {
                  new PointF(rect.Left, rect.Bottom),
                  new PointF(rect.Right, rect.Bottom),
                  new PointF(rect.Right, rect.Top),
                  new PointF(rect.Left, rect.Top)
                };
                points = CvInvoke.PerspectiveTransform(points, homography);
                Point[] points2 = Array.ConvertAll<PointF, Point>(points, Point.Round);
                VectorOfPoint vp = new VectorOfPoint(points2);
                CvInvoke.Polylines(result, vp, true, new MCvScalar(255, 0, 0, 255), 15);
            }

            imageBox1.Image = result;
        }
    }

效果图：