0

我正在使用 OpenCV4Android 2.4.11 版,并且正在检测相机检索到的帧中的任何矩形形状。如下图 1 所示,我在检测到的对象周围绘制了一个黑色轮廓,我想要做的是,获取绘制轮廓的所有坐标,该坐标仅以黑色绘制。我尝试的是,如下面的code_1所示,我得到最大轮廓和最大轮廓的索引,并将它们分别保存在“largestContour”和“largest_contour_index”中。然后,我使用

Imgproc.drawContours(mMatInputFrame, contours, largest_contour_index, new Scalar(0, 0, 0), 2, 8, hierachy, 0, new Point());

然后我将最大轮廓的点传递给 FindCorners 类,因为我想找到用黑色绘制的轮廓的特定坐标,如下所示:

this.mFindCorners = new FindCorners(largestContour.toArray());
double[] cords = this.mFindCorners.getCords();

以下代码行:

double[] cords = this.mFindCorners.getCords();

应该给我较小的 x 坐标、最小的 y 坐标、最大的 x 坐标和最大的 y 坐标。但是当我围绕从“this.mFindCorners.getCords();”得到的坐标画一个圆圈时 我在下面的 image_2 中得到了类似的东西,这只是 BoundingRect 的角落。

实际上我不希望 boundingRect 中的任何坐标我想访问在黑色检测到的对象周围绘制的轮廓的坐标

请让我知道如何获取轮廓本身的坐标?

代码_1

if (contours.size() > 0) {
    for (int i = 0; i < contours.size(); i++) {
        contour2f = new MatOfPoint2f(contours.get(i).toArray());
        approxDistance = Imgproc.arcLength(contour2f, true) * .01;//.02
        approxCurve = new MatOfPoint2f();
        Imgproc.approxPolyDP(contour2f, approxCurve, approxDistance, true);
        points = new MatOfPoint(approxCurve.toArray());

        double area = Math.abs(Imgproc.contourArea(points, true));
        if (points.total() >= 4 && area >= 40000 && area <= 200000) {
            if (area > largest_area) {
                largest_area = area;
                largest_contour_index = i;
                pointsOfLargestContour = points;
                largestContour = contours.get(i);
            }
        }
    }

    if (largest_area > 0) {
        Imgproc.drawContours(mMatInputFrame, contours, largest_contour_index, new Scalar(0, 0, 0), 2, 8, hierachy, 0, new Point());

        this.mFindCorners = new FindCorners(largestContour.toArray());
        double[] cords = this.mFindCorners.getCords();

        Core.circle(mMatInputFrame, new Point(cords[0], cords[1]), 10, new Scalar(255, 0, 0));
        Core.circle(mMatInputFrame, new Point(cords[2], cords[3]), 10, new Scalar(255, 255, 0));
    }

找角

public class FindCorners {
private final static String TAG = FragOpenCVCam.class.getSimpleName();

private ArrayList<Double> mlistXCords = null;
private ArrayList<Double> mlistYCords = null;
private double mSmallestX;
private double mSmallestY;
private double mLargestX;
private double mLargestY;
private double[] mCords = null;

public FindCorners(Point[] points) {
    this.mlistXCords = new ArrayList<>();
    this.mlistYCords = new ArrayList<>();
    this.mCords = new double[4];

    Log.d(TAG, "points.length: " + points.length);

    for (int i = 0; i < points.length; i++) {
        this.mlistXCords.add(points[i].x);
        this.mlistYCords.add(points[i].y);
    }

    //ascending
    Collections.sort(this.mlistXCords);
    Collections.sort(this.mlistYCords);

    this.mSmallestX = this.mlistXCords.get(0);
    this.mSmallestY = this.mlistYCords.get(0);
    this.mLargestX = this.mlistXCords.get(this.mlistXCords.size() - 1);
    this.mLargestY = this.mlistYCords.get(this.mlistYCords.size() - 1);

    this.mCords[0] = this.mSmallestX;
    this.mCords[1] = this.mSmallestY;
    this.mCords[2] = this.mLargestX;
    this.mCords[3] = this.mLargestY;
}

public double[] getCords() {
    return this.mCords;
}

}

图像_1

在此处输入图像描述

图像_2

在此处输入图像描述

更新 我不想拥有边界矩形的坐标,我想要的是黑色轮廓的确切坐标。如图 3 所示,我从代码中获得的坐标是红色和黄色圆圈所在的位置..但我正在寻找能够访问黑线“轮廓”的坐标,因此我可以在其上绘制一些圆圈,如图所示图像_3。绿色的点只是为了告诉你我想要的坐标。

图像_3

在此处输入图像描述

4

1 回答 1

0

您的问题是您已经分别对 x 和 y 进行了排序,很明显您的算法会找到红点和黄点。我可以建议以下算法:

int min_x=INF, min_x_index, min_y=1000, min_y_index;
int max_x=-1, max_x_index, max_y=-1, max_y_index;
 for (int i = 0; i < points.length; i++) 
    {
        if(points[i].x < min_x) { min_x = points[i].x; min_x_index = i; }
        if(points[i].x > max_x) { max_x = points[i].x; max_x_index = i; }
        if(points[i].y < min_y) { min_y = points[i].y; min_y_index = i; }
        if(points[i].y > max_y) { max_y = points[i].y; max_y_index = i; }
    }
Point corner1(points[min_x_index]);
Point corner2(points[min_y_index]);
Point corner3(points[max_x_index]);
Point corner4(points[max_y_index]);
于 2016-12-25T17:32:11.290 回答