我有点困惑与getPerspectiveTransform的参数，因为我看不到正确的图像。这是我的code。该original_image变量是包含一个方形物体（及其他​​），我要裁剪，并创建一个新的图像（像这样Android OpenCV中找到最大的正方形或长方形）。变量P1，P2，P3和P4是最大正方形/矩形的图像中的角的坐标。 p1的左上方，p2为右上，p3的是右下，和p4是左下（顺时针分配）。

I am a bit confused with the parameters of getPerspectiveTransform as I cannot see a proper image. Here is my code. The original_image variable is the image that contains a square object (and some others) that I want to crop and create a new image (something like this Android OpenCV Find Largest Square or Rectangle). The variables p1, p2, p3, and p4 are the coordinates of the corners of the largest square/rectangle in the image. p1 is the upper left, p2 is the upper right, p3 is the lower right, and p4 is the lower left (clockwise assigning).

    Mat src = new Mat(4,1,CvType.CV_32FC2);
    src.put((int)p1.y,(int)p1.x, (int)p2.y,(int)p2.x, (int)p4.y,(int)p4.x, (int)p3.y,(int)p3.x);
    Mat dst = new Mat(4,1,CvType.CV_32FC2);
    dst.put(0,0, 0,original_image.width(), original_image.height(),original_image.width(), original_image.height(),0);

    Mat perspectiveTransform = Imgproc.getPerspectiveTransform(src, dst);
    Mat cropped_image = original_image.clone();
    Imgproc.warpPerspective(untouched, cropped_image, perspectiveTransform, new Size(512,512));

当我尝试显示cropped_image，我得到一个我不知道它是什么的形象。我觉得我在getPerspectiveTransform（）的参数是不正确（或它）。请帮忙。谢谢！

When I try to display cropped_image, I get a "I don't know what it is" image. I think my parameters in getPerspectiveTransform() are incorrect (or is it). Please help. Thanks!

更新：当我调试我的code，我发现我的正方形/长方形的边是不正确的，也有些是完全正确的，除了P4。这是我的code键检测图像中的正方形或长方形的边缘。我的图像是全黑的，除了最大的正方形/长方形具有白色轮廓的轮廓。

Update: When I debugged my code, I found out that the edges of my square/rectangle are incorrect, well some are quite right except for p4. This is my code to detect the edges of the square or rectangle in the image. My image is all black except for the contour of the largest square/rectangle which has a white outline.

    //we will find the edges of the new_image (corners of the square/rectangle)
    Point p1 = new Point(10000, 10000); //upper left; minX && minY
    Point p2 = new Point(0, 10000); //upper right; maxX && minY
    Point p3 = new Point(0, 0); //lower right; maxX && maxY
    Point p4 = new Point(10000, 0); //lower left; minX && maxY
    double[] temp_pixel_color;
    for (int x=0; x<new_image.rows(); x++) {
        for (int y=0; y<new_image.cols(); y++) {
            temp_pixel_color = new_image.get(x, y); //we have a black and white image so we only have one color channel
            if (temp_pixel_color[0] > 200) { //we found a white pixel
                if (x<=p1.x && y<=p1.y) { //for p1, minX && minY
                    p1.x = x;
                    p1.y = y;
                }
                else if (x>=p2.x && y<=p2.y) { //for p2, maxX && minY
                    p2.x = x;
                    p2.y = y;
                }
                else if (x>=p3.x && y>=p3.y) { //for p3, maxX && maxY
                    p3.x = x;
                    p3.y = y;
                }
                else if (x<=(int)p4.x && y>=(int)p4.y) { //for p4, minX && maxY
                    p4.x = x;
                    p4.y = y;
                }
            }
        }
    }

下面是我的样本图像。忽略彩色圆圈，因为他们得出被检测的边缘后：

Here is my sample image. ignore the colored circles as they are drawn after the edges are detected:

更新：2013年7月16日 我能察觉现在用的最多4个尖的轮廓，只有approxCurve角落。这是我的code：

Update: July 16, 2013 I can detect the corners now using only the approxCurve of the maximum 4-pointed contour. Here is my code:

private Mat findLargestRectangle(Mat original_image) {
    Mat imgSource = original_image;
    //Mat untouched = original_image.clone();

    //convert the image to black and white
    Imgproc.cvtColor(imgSource, imgSource, Imgproc.COLOR_BGR2GRAY);

    //convert the image to black and white does (8 bit)
    Imgproc.Canny(imgSource, imgSource, 50, 50);

    //apply gaussian blur to smoothen lines of dots
    Imgproc.GaussianBlur(imgSource, imgSource, new Size(5, 5), 5);      

    //find the contours
    List<MatOfPoint> contours = new ArrayList<MatOfPoint>();
    Imgproc.findContours(imgSource, contours, new Mat(), Imgproc.RETR_LIST, Imgproc.CHAIN_APPROX_SIMPLE);

    double maxArea = -1;
    int maxAreaIdx = -1;
    MatOfPoint temp_contour = contours.get(0); //the largest is at the index 0 for starting point
    MatOfPoint2f approxCurve = new MatOfPoint2f();
    MatOfPoint2f maxCurve = new MatOfPoint2f();
    List<MatOfPoint> largest_contours = new ArrayList<MatOfPoint>();
    for (int idx = 0; idx < contours.size(); idx++) {
        temp_contour = contours.get(idx);
        double contourarea = Imgproc.contourArea(temp_contour);
        //compare this contour to the previous largest contour found
        if (contourarea > maxArea) {
            //check if this contour is a square
            MatOfPoint2f new_mat = new MatOfPoint2f( temp_contour.toArray() );
            int contourSize = (int)temp_contour.total();
            Imgproc.approxPolyDP(new_mat, approxCurve, contourSize*0.05, true);
            if (approxCurve.total() == 4) {
                maxCurve = approxCurve;
                maxArea = contourarea;
                maxAreaIdx = idx;
                largest_contours.add(temp_contour);
            }
        }
    }

    //create the new image here using the largest detected square
    Mat new_image = new Mat(imgSource.size(), CvType.CV_8U); //we will create a new black blank image with the largest contour
    Imgproc.cvtColor(new_image, new_image, Imgproc.COLOR_BayerBG2RGB);
    Imgproc.drawContours(new_image, contours, maxAreaIdx, new Scalar(255, 255, 255), 1); //will draw the largest square/rectangle

    double temp_double[] = maxCurve.get(0, 0);
    Point p1 = new Point(temp_double[0], temp_double[1]);
    Core.circle(new_image, new Point(p1.x, p1.y), 20, new Scalar(255, 0, 0), 5); //p1 is colored red
    String temp_string = "Point 1: (" + p1.x + ", " + p1.y + ")";

    temp_double = maxCurve.get(1, 0);
    Point p2 = new Point(temp_double[0], temp_double[1]);
    Core.circle(new_image, new Point(p2.x, p2.y), 20, new Scalar(0, 255, 0), 5); //p2 is colored green
    temp_string += "\nPoint 2: (" + p2.x + ", " + p2.y + ")";

    temp_double = maxCurve.get(2, 0);       
    Point p3 = new Point(temp_double[0], temp_double[1]);
    Core.circle(new_image, new Point(p3.x, p3.y), 20, new Scalar(0, 0, 255), 5); //p3 is colored blue
    temp_string += "\nPoint 3: (" + p3.x + ", " + p3.y + ")";

    temp_double = maxCurve.get(3, 0);
    Point p4 = new Point(temp_double[0], temp_double[1]);
    Core.circle(new_image, new Point(p4.x, p4.y), 20, new Scalar(0, 255, 255), 5); //p1 is colored violet
    temp_string += "\nPoint 4: (" + p4.x + ", " + p4.y + ")";

    TextView temp_text = (TextView)findViewById(R.id.temp_text);
    temp_text.setText(temp_string);

    return new_image;
}

下面是示例结果图像：

我已经引起各界的正方形/长方形的四个角和我还添加了一个TextView来显示所有的四点意见。

I have drawn circles for the corners of the square/rectangle and I also added a textview to display all the four points.

推荐答案

这为我工作。在src_mat.put你应该有0,0处，然后再为坐标浮点值。

This worked for me. in the src_mat.put you should have 0,0 at first and then the float values for the coordinates.

    Mat mat=Highgui.imread("inputImage.jpg");
    Mat src_mat=new Mat(4,1,CvType.CV_32FC2);
    Mat dst_mat=new Mat(4,1,CvType.CV_32FC2);


    src_mat.put(0,0,407.0,74.0,1606.0,74.0,420.0,2589.0,1698.0,2589.0);
    dst_mat.put(0,0,0.0,0.0,1600.0,0.0, 0.0,2500.0,1600.0,2500.0);
    Mat perspectiveTransform=Imgproc.getPerspectiveTransform(src_mat, dst_mat);

    Mat dst=mat.clone();

    Imgproc.warpPerspective(mat, dst, perspectiveTransform, new Size(1600,2500));
    Highgui.imwrite("resultImage.jpg", dst);