标签：
---恢复内容开始---
最近没什么事，公司领导希望我们了解图像处理，所以学习了以下！由于我不太会C++，只能试着用C#编写代码！但是网上关于emgu cv的资料少之又少，而且很多还是英文的，而且讲的不详细。所以慢慢琢磨。写了个c#定位车牌的代码，不过效果不是很理想。参考了c++高手的代码！
思路就是1.灰度化，竖向边缘检测
2.自适应二值化处理
3.形态学处理（膨胀和腐蚀）
4.轮廓查找与筛选
代码如下：
Image<Bgr, Byte> simage = img; //new Image<Bgr, byte>("license-plate.jpg"); //Image<Bgr, Byte> simage = sizeimage.Resize(400, 300, Emgu.CV.CvEnum.INTER.CV_INTER_NN); Image<Gray, Byte> GrayImg = new Image<Gray, Byte>(simage.Width, simage.Height); IntPtr GrayImg1 = CvInvoke.cvCreateImage(simage.Size, Emgu.CV.CvEnum.IPL_DEPTH.IPL_DEPTH_8U, 1); //灰度化 CvInvoke.cvCvtColor(simage.Ptr, GrayImg1, Emgu.CV.CvEnum.COLOR_CONVERSION.BGR2GRAY); //首先创建一张16深度有符号的图像区域 IntPtr Sobel = CvInvoke.cvCreateImage(simage.Size, Emgu.CV.CvEnum.IPL_DEPTH.IPL_DEPTH_16S, 1); //X方向的Sobel算子检测 CvInvoke.cvSobel(GrayImg1, Sobel, 2, 0, 3); IntPtr temp = CvInvoke.cvCreateImage(simage.Size, Emgu.CV.CvEnum.IPL_DEPTH.IPL_DEPTH_8U, 1); CvInvoke.cvConvertScale(Sobel, temp, 0.00390625, 0); ////int it = ComputeThresholdValue(GrayImg.ToBitmap()); ////二值化处理 ////Image<Gray, Byte> dest = GrayImg.ThresholdBinary(new Gray(it), new Gray(255)); Image<Gray, Byte> dest = new Image<Gray, Byte>(simage.Width, simage.Height); //二值化处理 CvInvoke.cvThreshold(temp, dest, 0, 255, Emgu.CV.CvEnum.THRESH.CV_THRESH_OTSU); IntPtr temp1 = CvInvoke.cvCreateImage(simage.Size, Emgu.CV.CvEnum.IPL_DEPTH.IPL_DEPTH_8U, 1); Image<Gray, Byte> dest1 = new Image<Gray, Byte>(simage.Width, simage.Height); CvInvoke.cvCreateStructuringElementEx(3, 1, 1, 0, Emgu.CV.CvEnum.CV_ELEMENT_SHAPE.CV_SHAPE_RECT, temp1); CvInvoke.cvDilate(dest, dest1, temp1, 6); CvInvoke.cvErode(dest1, dest1, temp1, 7); CvInvoke.cvDilate(dest1, dest1, temp1, 1); CvInvoke.cvCreateStructuringElementEx(1, 3, 0, 1, Emgu.CV.CvEnum.CV_ELEMENT_SHAPE.CV_SHAPE_RECT, temp1); CvInvoke.cvErode(dest1, dest1, temp1, 2); CvInvoke.cvDilate(dest1, dest1, temp1, 2); IntPtr dst = CvInvoke.cvCreateImage(simage.Size, Emgu.CV.CvEnum.IPL_DEPTH.IPL_DEPTH_8U, 3); CvInvoke.cvZero(dst); //dest.Dilate(10); //dest.Erode(5); using (MemStorage stor = new MemStorage()) { Contour<Point> contours = dest1.FindContours( Emgu.CV.CvEnum.CHAIN_APPROX_METHOD.CV_CHAIN_APPROX_SIMPLE, Emgu.CV.CvEnum.RETR_TYPE.CV_RETR_CCOMP, stor); for (; contours != null; contours = contours.HNext) { Rectangle box = contours.BoundingRectangle; Image<Bgr, Byte> test = simage.CopyBlank(); test.SetValue(255.0); double whRatio = (double)box.Width / box.Height; int area = (int)box.Width * box.Height; if (area > 1000 && area<10000) { if ((3.0 < whRatio && whRatio < 6.0)) { test.Draw(box, new Bgr(Color.Red), 2); simage.Draw(box, new Bgr(Color.Red), 2); CvInvoke.cvRectangle(simage, new Point(box.X, box.Y), new Point(box.X + box.Width, box.Y + box.Height), new MCvScalar(255, 0, 0), 1, Emgu.CV.CvEnum.LINE_TYPE.EIGHT_CONNECTED, 0); //CvInvoke.cvNamedWindow("dst"); //CvInvoke.cvShowImage("dst", dst); imageBox1.Image = simage; } } } }
还是有一些细节没处理好啊
标签：
原文地址：http://www.cnblogs.com/defly/p/4269241.html
