特征检测后的OpenCV的Android提取匹配垫 [英] OpenCV Android extract match Mat after feature detection
问题描述
我想提取匹配我的参考图像的图像的一部分。
我尝试使用Calib3d.findHomography方法来转换图像。
当我这样做，我使用Imgproc.warpPerspective做改造，但没有很好的效果。难道我在这里错过了什么？
我需要做一些与perspectiveTransform？我曾经试过，但没有到目前为止，任何侥幸。
下面是我的findSceneCorners方式：
私人无效findSceneCorners（太SRC）{
mFeatureDetector.detect（SRC，mSceneKeypoints）;
mDescriptorExtractor.compute（SRC，mSceneKeypoints，mSceneDescriptors）;
mDescriptorMatcher.match（mSceneDescriptors，mReferenceDescriptors，mMatches）; 清单＆LT; DMatch＆GT; matchesList = mMatches.toList（）;
如果（matchesList.size（）4;）{
//有太多几场比赛找到对应性。
返回;
} 清单＆LT;关键点＆GT; referenceKeypointsList =
mReferenceKeypoints.toList（）;
清单＆LT;关键点＆GT; sceneKeypointsList =
mSceneKeypoints.toList（）; //计算最大值和关键点之间的距离最小。
双maxDist = 0.0;
双minDist = Double.MAX_VALUE;
对于（DMatch匹配：matchesList）{
双DIST = match.distance;
如果（DIST＆LT; minDist）{
minDist = DIST;
}
如果（DIST＆GT; maxDist）{
maxDist = DIST;
}
} //为minDist的阈值是主观选择
//根据测试。该单元是不相关的像素
//距离;它关系到失败的测试数
//为匹配的描述符之间的相似性。
如果（minDist＆GT; 50.0）{
//目标是完全丧失。
//丢弃任何previously发现角落。
mSceneCorners.create（0，0，mSceneCorners.type（））;
返回;
}否则如果（minDist＆20.0）{
//目标丢失，但也许它仍然接近。
//保留任何previously发现角落。
返回;
}
//确定基于匹配距离好的关键点。
ArrayList的＆LT;点和GT; goodReferencePointsList =
新的ArrayList＆LT;点和GT;（）;
ArrayList的＆LT;点和GT; goodScenePointsList =
新的ArrayList＆LT;点和GT;（）; 双maxGoodMatchDist = 1.75 * minDist;
对于（DMatch匹配：matchesList）{
如果（match.distance＆LT; maxGoodMatchDist）{
goodReferencePointsList.add（
referenceKeypointsList.get（match.trainIdx）.PT）;
goodScenePointsList.add（
sceneKeypointsList.get（match.queryIdx）.PT）;
}
} 如果（goodReferencePointsList.size（）4; ||
goodScenePointsList.size（）＆下; 4）{
//有找到单应太少好点。
返回;
} Log.i（TAG，匹配找到）; MatOfPoint2f goodReferencePoints =新MatOfPoint2f（）;
goodReferencePoints.fromList（goodReferencePointsList）; MatOfPoint2f goodScenePoints =新MatOfPoint2f（）;
goodScenePoints.fromList（goodScenePointsList）; 单应= Calib3d.findHomography（goodReferencePoints，goodScenePoints）; 垫四=新垫（mReferenceImage.size（），CvType.CV_32F）;
Imgproc.warpPerspective（SRC，四，单应，quad.size（））;
objectDetectedListener.objectDetected（四）}
我认为你应该使用 WARP_INVERSE_MAP 作为的标志 warpPerspective 为： Imgproc.warpPerspective（SRC，四，单应，quad.size（），WARP_INVERSE_MAP）;
我没有使用完全相同的code，但只是单应后的部分，我已经看到，图像在镜扭曲，不是因为我们想（使用更大的显示图像，看看究竟在那里） 。事实上，你贴出来，与10卡的页面上，则使用该标志，抱歉，我没有想过提到这更早的。
I want to extract the part of the image that matches my reference image. I try to transform the image using the Calib3d.findHomography method. When i have done this i use Imgproc.warpPerspective to do the transformation but with no good results. Do i miss something here? Do i need to do something with perspectiveTransform? I have tried this but without any luck so far.
Here is my findSceneCorners method:
private void findSceneCorners(Mat src) {
mFeatureDetector.detect(src, mSceneKeypoints);
mDescriptorExtractor.compute(src, mSceneKeypoints, mSceneDescriptors);
mDescriptorMatcher.match(mSceneDescriptors, mReferenceDescriptors, mMatches);
List<DMatch> matchesList = mMatches.toList();
if (matchesList.size() < 4) {
// There are too few matches to find the homography.
return;
}
List<KeyPoint> referenceKeypointsList =
mReferenceKeypoints.toList();
List<KeyPoint> sceneKeypointsList =
mSceneKeypoints.toList();
// Calculate the max and min distances between keypoints.
double maxDist = 0.0;
double minDist = Double.MAX_VALUE;
for(DMatch match : matchesList) {
double dist = match.distance;
if (dist < minDist) {
minDist = dist;
}
if (dist > maxDist) {
maxDist = dist;
}
}
// The thresholds for minDist are chosen subjectively
// based on testing. The unit is not related to pixel
// distances; it is related to the number of failed tests
// for similarity between the matched descriptors.
if (minDist > 50.0) {
// The target is completely lost.
// Discard any previously found corners.
mSceneCorners.create(0, 0, mSceneCorners.type());
return;
} else if (minDist > 20.0) {
// The target is lost but maybe it is still close.
// Keep any previously found corners.
return;
}
// Identify "good" keypoints based on match distance.
ArrayList<Point> goodReferencePointsList =
new ArrayList<Point>();
ArrayList<Point> goodScenePointsList =
new ArrayList<Point>();
double maxGoodMatchDist = 1.75 * minDist;
for(DMatch match : matchesList) {
if (match.distance < maxGoodMatchDist) {
goodReferencePointsList.add(
referenceKeypointsList.get(match.trainIdx).pt);
goodScenePointsList.add(
sceneKeypointsList.get(match.queryIdx).pt);
}
}
if (goodReferencePointsList.size() < 4 ||
goodScenePointsList.size() < 4) {
// There are too few good points to find the homography.
return;
}
Log.i(TAG, "Match found");
MatOfPoint2f goodReferencePoints = new MatOfPoint2f();
goodReferencePoints.fromList(goodReferencePointsList);
MatOfPoint2f goodScenePoints = new MatOfPoint2f();
goodScenePoints.fromList(goodScenePointsList);
homography = Calib3d.findHomography(goodReferencePoints, goodScenePoints);
Mat quad = new Mat(mReferenceImage.size(), CvType.CV_32F);
Imgproc.warpPerspective(src, quad, homography, quad.size());
objectDetectedListener.objectDetected(quad);
}
I think you should use WARP_INVERSE_MAP as the flag of warpPerspective as: Imgproc.warpPerspective(src, quad, homography, quad.size(),WARP_INVERSE_MAP);.
I have not used exactly your code, but just the part after homography, and i have seen that the image was warped in the mirror, not as we wanted (use a bigger display image to see exactly what's there). In fact, on the page you posted, with the 10-card, it is used that flag, sorry that I have not thought of mentioning this earlier.
这篇关于特征检测后的OpenCV的Android提取匹配垫的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！
