org.bytedeco.javacv.GeometricCalibrator Maven / Gradle / Ivy
/*
* Copyright (C) 2009-2011 Samuel Audet
*
* Licensed either under the Apache License, Version 2.0, or (at your option)
* under the terms of the GNU General Public License as published by
* the Free Software Foundation (subject to the "Classpath" exception),
* either version 2, or any later version (collectively, the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
* http://www.gnu.org/licenses/
* http://www.gnu.org/software/classpath/license.html
*
* or as provided in the LICENSE.txt file that accompanied this code.
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.bytedeco.javacv;
import java.nio.FloatBuffer;
import java.nio.IntBuffer;
import java.util.Arrays;
import java.util.Iterator;
import java.util.LinkedList;
import org.bytedeco.javacv.ProjectiveDevice.CalibrationSettings;
import org.bytedeco.opencv.opencv_calib3d.*;
import org.bytedeco.opencv.opencv_core.*;
import org.bytedeco.opencv.opencv_imgproc.*;
import static org.bytedeco.opencv.global.opencv_calib3d.*;
import static org.bytedeco.opencv.global.opencv_core.*;
import static org.bytedeco.opencv.global.opencv_imgproc.*;
/**
*
* @author Samuel Audet
*/
public class GeometricCalibrator {
public GeometricCalibrator(Settings settings, MarkerDetector.Settings detectorSettings,
MarkedPlane markedPlane, ProjectiveDevice projectiveDevice) {
this.settings = settings;
this.markerDetector = new MarkerDetector(detectorSettings);
this.markedPlane = markedPlane;
this.projectiveDevice = projectiveDevice;
cvSetIdentity(prevWarp);
cvSetIdentity(lastWarp);
if (markedPlane != null) {
int w = markedPlane.getImage().width();
int h = markedPlane.getImage().height();
warpSrcPts.put(0.0, 0.0, w, 0.0, w, h, 0.0, h);
}
}
public static class Settings extends BaseChildSettings {
double detectedBoardMin = 0.5;
double patternSteadySize = 0.005;
double patternMovedSize = 0.05;
public double getDetectedBoardMin() {
return detectedBoardMin;
}
public void setDetectedBoardMin(double detectedBoardMin) {
this.detectedBoardMin = detectedBoardMin;
}
public double getPatternSteadySize() {
return patternSteadySize;
}
public void setPatternSteadySize(double patternSteadySize) {
this.patternSteadySize = patternSteadySize;
}
public double getPatternMovedSize() {
return patternMovedSize;
}
public void setPatternMovedSize(double patternMovedSize) {
this.patternMovedSize = patternMovedSize;
}
}
private Settings settings;
MarkerDetector markerDetector;
private MarkedPlane markedPlane;
private ProjectiveDevice projectiveDevice;
private LinkedList allObjectMarkers = new LinkedList();
private LinkedList allImageMarkers = new LinkedList();
private IplImage tempImage = null;
private Marker[] lastDetectedMarkers = null;
private CvMat warp = CvMat.create(3, 3);
private CvMat prevWarp = CvMat.create(3, 3);
private CvMat lastWarp = CvMat.create(3, 3);
private CvMat warpSrcPts = CvMat.create(1, 4, CV_64F, 2);
private CvMat warpDstPts = CvMat.create(1, 4, CV_64F, 2);
private CvMat tempPts = CvMat.create(1, 4, CV_64F, 2);
public MarkerDetector getMarkerDetector() {
return markerDetector;
}
public MarkedPlane getMarkedPlane() {
return markedPlane;
}
public ProjectiveDevice getProjectiveDevice() {
return projectiveDevice;
}
public LinkedList getAllObjectMarkers() {
return allObjectMarkers;
}
public void setAllObjectMarkers(LinkedList allObjectMarkers) {
this.allObjectMarkers = allObjectMarkers;
}
public LinkedList getAllImageMarkers() {
return allImageMarkers;
}
public void setAllImageMarkers(LinkedList allImageMarkers) {
this.allImageMarkers = allImageMarkers;
}
public Marker[] processImage(IplImage image) {
projectiveDevice.imageWidth = image.width();
projectiveDevice.imageHeight = image.height();
final boolean whiteMarkers = markedPlane.getForegroundColor().magnitude() >
markedPlane.getBackgroundColor().magnitude();
if (image.depth() > 8) {
if (tempImage == null ||
tempImage.width() != image.width() ||
tempImage.height() != image.height()) {
tempImage = (IplImage)IplImage.create(image.width(), image.height(),
IPL_DEPTH_8U, 1, image.origin());
}
cvConvertScale(image, tempImage, 1.0/(1<<8), 0);
lastDetectedMarkers = markerDetector.detect(tempImage, whiteMarkers);
} else {
lastDetectedMarkers = markerDetector.detect(image, whiteMarkers);
}
// First, check if we detected enough markers
if (lastDetectedMarkers.length < markedPlane.getMarkers().length*settings.detectedBoardMin) {
return null;
}
// then check by how much the corners of the calibration board moved
markedPlane.getTotalWarp(lastDetectedMarkers, warp);
cvPerspectiveTransform (warpSrcPts, warpDstPts, warp);
cvPerspectiveTransform (warpSrcPts, tempPts, prevWarp);
double rmsePrev = cvNorm(warpDstPts, tempPts);
cvPerspectiveTransform (warpSrcPts, tempPts, lastWarp);
double rmseLast = cvNorm(warpDstPts, tempPts);
//System.out.println("rmsePrev = " + rmsePrev + " rmseLast = " + rmseLast);
// save warp for next iteration...
cvCopy(warp, prevWarp);
// send upstream our recommendation for addition or not of these markers...
int imageSize = (image.width()+image.height())/2;
if (rmsePrev < settings.patternSteadySize*imageSize &&
rmseLast > settings.patternMovedSize*imageSize) {
return lastDetectedMarkers;
} else {
return null;
}
}
public void drawMarkers(IplImage image) {
markerDetector.draw(image, lastDetectedMarkers);
}
public void addMarkers() {
addMarkers(markedPlane.getMarkers(), lastDetectedMarkers);
}
public void addMarkers(Marker[] imageMarkers) {
addMarkers(markedPlane.getMarkers(), imageMarkers);
}
public void addMarkers(Marker[] objectMarkers, Marker[] imageMarkers) {
// add only matching markers...
int maxLength = Math.min(objectMarkers.length, imageMarkers.length);
Marker[] om = new Marker[maxLength];
Marker[] im = new Marker[maxLength];
int i = 0;
for (Marker m1 : objectMarkers) {
for (Marker m2 : imageMarkers) {
if (m1.id == m2.id) {
om[i] = m1;
im[i] = m2;
i++;
break;
}
}
}
if (i < maxLength) {
om = Arrays.copyOf(om, i);
im = Arrays.copyOf(im, i);
}
allObjectMarkers.add(om);
allImageMarkers.add(im);
// we added the detected markers, so save last computed warp too...
cvCopy(prevWarp, lastWarp);
}
public int getImageCount() {
assert(allObjectMarkers.size() == allImageMarkers.size());
return allObjectMarkers.size();
}
private Point3fVectorVector getObjectPoints(CvMat points, CvMat counts) {
FloatBuffer pointsBuf = points.getFloatBuffer();
IntBuffer countsBuf = counts.getIntBuffer();
int n = counts.length();
Point3fVectorVector vectors = new Point3fVectorVector(n);
for (int i = 0; i < n; i++) {
int m = countsBuf.get();
Point3fVector vector = new Point3fVector(m);
for (int j = 0; j < m; j++) {
vector.put(j, new Point3f(pointsBuf.get(), pointsBuf.get(), pointsBuf.get()));
}
vectors.put(i, vector);
}
return vectors;
}
private Point2fVectorVector getImagePoints(CvMat points, CvMat counts) {
FloatBuffer pointsBuf = points.getFloatBuffer();
IntBuffer countsBuf = counts.getIntBuffer();
int n = counts.length();
Point2fVectorVector vectors = new Point2fVectorVector(n);
for (int i = 0; i < n; i++) {
int m = countsBuf.get();
Point2fVector vector = new Point2fVector(m);
for (int j = 0; j < m; j++) {
vector.put(j, new Point2f(pointsBuf.get(), pointsBuf.get()));
}
vectors.put(i, vector);
}
return vectors;
}
private CvMat[] getPoints(boolean useCenters) {
// fill up pointCounts, objectPoints and imagePoints, with data from
// srcMarkers and dstMarkers
assert(allObjectMarkers.size() == allImageMarkers.size());
Iterator i1 = allObjectMarkers.iterator(),
i2 = allImageMarkers.iterator();
CvMat pointCounts = CvMat.create(1, allImageMarkers.size(), CV_32S, 1);
IntBuffer pointCountsBuf = pointCounts.getIntBuffer();
int totalPointCount = 0;
while (i1.hasNext() && i2.hasNext()) {
Marker[] m1 = i1.next(),
m2 = i2.next();
assert(m1.length == m2.length);
int n = m1.length*(useCenters ? 1 : 4);
pointCountsBuf.put(n);
totalPointCount += n;
}
i1 = allObjectMarkers.iterator();
i2 = allImageMarkers.iterator();
CvMat objectPoints = CvMat.create(1, totalPointCount, CV_32F, 3);
CvMat imagePoints = CvMat.create(1, totalPointCount, CV_32F, 2);
FloatBuffer objectPointsBuf = objectPoints.getFloatBuffer();
FloatBuffer imagePointsBuf = imagePoints.getFloatBuffer();
while (i1.hasNext() && i2.hasNext()) {
Marker[] m1 = i1.next(),
m2 = i2.next();
for (int j = 0; j < m1.length; j++) {
if (useCenters) {
double[] c1 = m1[j].getCenter();
objectPointsBuf.put((float)c1[0]);
objectPointsBuf.put((float)c1[1]);
objectPointsBuf.put(0);
double[] c2 = m2[j].getCenter();
imagePointsBuf.put((float)c2[0]);
imagePointsBuf.put((float)c2[1]);
} else { // use corners...
for (int k = 0; k < 4; k++) {
objectPointsBuf.put((float)m1[j].corners[2*k ]);
objectPointsBuf.put((float)m1[j].corners[2*k + 1]);
objectPointsBuf.put(0);
imagePointsBuf.put((float)m2[j].corners[2*k ]);
imagePointsBuf.put((float)m2[j].corners[2*k + 1]);
}
}
}
}
return new CvMat[] { objectPoints, imagePoints, pointCounts };
}
public static double[] computeReprojectionError(CvMat object_points,
CvMat image_points, CvMat point_counts, CvMat camera_matrix,
CvMat dist_coeffs, CvMat rot_vects, CvMat trans_vects,
CvMat per_view_errors ) {
CvMat image_points2 = CvMat.create(image_points.rows(),
image_points.cols(), image_points.type());
int i, image_count = rot_vects.rows(), points_so_far = 0;
double total_err = 0, max_err = 0, err;
CvMat object_points_i = new CvMat(),
image_points_i = new CvMat(),
image_points2_i = new CvMat();
IntBuffer point_counts_buf = point_counts.getIntBuffer();
CvMat rot_vect = new CvMat(), trans_vect = new CvMat();
for (i = 0; i < image_count; i++) {
object_points_i.reset();
image_points_i .reset();
image_points2_i.reset();
int point_count = point_counts_buf.get(i);
cvGetCols(object_points, object_points_i,
points_so_far, points_so_far + point_count);
cvGetCols(image_points, image_points_i,
points_so_far, points_so_far + point_count);
cvGetCols(image_points2, image_points2_i,
points_so_far, points_so_far + point_count);
points_so_far += point_count;
cvGetRows(rot_vects, rot_vect, i, i+1, 1);
cvGetRows(trans_vects, trans_vect, i, i+1, 1);
projectPoints(cvarrToMat(object_points_i), cvarrToMat(rot_vect), cvarrToMat(trans_vect),
cvarrToMat(camera_matrix), cvarrToMat(dist_coeffs), cvarrToMat(image_points2_i));
err = cvNorm(image_points_i, image_points2_i);
err *= err;
if (per_view_errors != null)
per_view_errors.put(i, Math.sqrt(err/point_count));
total_err += err;
for (int j = 0; j < point_count; j++) {
double x1 = image_points_i .get(0, j, 0);
double y1 = image_points_i .get(0, j, 1);
double x2 = image_points2_i.get(0, j, 0);
double y2 = image_points2_i.get(0, j, 1);
double dx = x1-x2;
double dy = y1-y2;
err = Math.sqrt(dx*dx + dy*dy);
if (err > max_err) {
max_err = err;
}
}
}
return new double[] { Math.sqrt(total_err/points_so_far), max_err };
}
public double[] calibrate(boolean useCenters) {
ProjectiveDevice d = projectiveDevice;
CalibrationSettings dsettings = (CalibrationSettings)d.getSettings();
if (d.cameraMatrix == null) {
d.cameraMatrix = CvMat.create(3, 3);
cvSetZero(d.cameraMatrix);
if ((dsettings.flags & CV_CALIB_FIX_ASPECT_RATIO) != 0) {
d.cameraMatrix.put(0, dsettings.initAspectRatio);
d.cameraMatrix.put(4, 1.);
}
}
int kn = dsettings.isFixK3() ? 4 : 5;
if (dsettings.isRationalModel() && !dsettings.isFixK4() &&
!dsettings.isFixK4() && !dsettings.isFixK5()) {
kn = 8;
}
if (d.distortionCoeffs == null || d.distortionCoeffs.cols() != kn) {
d.distortionCoeffs = CvMat.create(1, kn);
cvSetZero(d.distortionCoeffs);
}
CvMat rotVects = new CvMat(), transVects = new CvMat();
d.extrParams = CvMat.create(allImageMarkers.size(), 6);
cvGetCols(d.extrParams, rotVects, 0, 3);
cvGetCols(d.extrParams, transVects, 3, 6);
CvMat[] points = getPoints(useCenters);
MatVector rvecs = new MatVector();
MatVector tvecs = new MatVector();
Mat distortionCoeffs = new Mat();
calibrateCamera(getObjectPoints(points[0], points[2]), getImagePoints(points[1], points[2]),
new Size(d.imageWidth, d.imageHeight),
cvarrToMat(d.cameraMatrix), distortionCoeffs,
rvecs, tvecs, dsettings.flags,
new TermCriteria(CV_TERMCRIT_ITER | CV_TERMCRIT_EPS, 30, JavaCV.DBL_EPSILON));
int n = (int)rvecs.size();
CvMat row = new CvMat();
for (int i = 0; i < n; i++) {
cvTranspose(cvMat(rvecs.get(i)), cvGetRow(rotVects, row, i));
cvTranspose(cvMat(tvecs.get(i)), cvGetRow(transVects, row, i));
}
d.distortionCoeffs = cvMat(distortionCoeffs).clone();
if (cvCheckArr(d.cameraMatrix, CV_CHECK_QUIET, 0, 0) != 0 &&
cvCheckArr(d.distortionCoeffs, CV_CHECK_QUIET, 0, 0) != 0 &&
cvCheckArr(d.extrParams, CV_CHECK_QUIET, 0, 0) != 0) {
d.reprojErrs = CvMat.create(1, allImageMarkers.size());
double[] err = computeReprojectionError(points[0], points[1], points[2],
d.cameraMatrix, d.distortionCoeffs, rotVects, transVects, d.reprojErrs);
d.avgReprojErr = err[0];
d.maxReprojErr = err[1];
// d.nominalDistance = d.getNominalDistance(markedPlane);
return err;
} else {
d.cameraMatrix = null;
d.avgReprojErr = -1;
d.maxReprojErr = -1;
return null;
}
}
public static double[] computeStereoError(CvMat imagePoints1, CvMat imagePoints2,
CvMat M1, CvMat D1, CvMat M2, CvMat D2, CvMat F) {
// CALIBRATION QUALITY CHECK
// because the output fundamental matrix implicitly
// includes all the output information,
// we can check the quality of calibration using the
// epipolar geometry constraint: m2^t*F*m1=0
int N = imagePoints1.cols();
CvMat L1 = CvMat.create(1, N, CV_32F, 3);
CvMat L2 = CvMat.create(1, N, CV_32F, 3);
//Always work in undistorted space
undistortPoints(cvarrToMat(imagePoints1), cvarrToMat(imagePoints1), cvarrToMat(M1), cvarrToMat(D1), null, cvarrToMat(M1));
undistortPoints(cvarrToMat(imagePoints2), cvarrToMat(imagePoints2), cvarrToMat(M2), cvarrToMat(D2), null, cvarrToMat(M2));
//imagePoints1.put(d1.undistort(imagePoints1.get()));
//imagePoints2.put(d2.undistort(imagePoints2.get()));
computeCorrespondEpilines(cvarrToMat(imagePoints1), 1, cvarrToMat(F), cvarrToMat(L1));
computeCorrespondEpilines(cvarrToMat(imagePoints2), 2, cvarrToMat(F), cvarrToMat(L2));
double avgErr = 0, maxErr = 0;
CvMat p1 = imagePoints1, p2 = imagePoints2;
for(int i = 0; i < N; i++ ) {
double e1 = p1.get(0, i, 0)*L2.get(0, i, 0) +
p1.get(0, i, 1)*L2.get(0, i, 1) + L2.get(0, i, 2);
double e2 = p2.get(0, i, 0)*L1.get(0, i, 0) +
p2.get(0, i, 1)*L1.get(0, i, 1) + L1.get(0, i, 2);
double err = e1*e1 + e2*e2;
avgErr += err;
err = Math.sqrt(err);
if (err > maxErr) {
maxErr = err;
}
}
return new double[] { Math.sqrt(avgErr/N), maxErr };
}
public double[] calibrateStereo(boolean useCenters, GeometricCalibrator peer) {
ProjectiveDevice d = projectiveDevice;
ProjectiveDevice dp = peer.projectiveDevice;
CalibrationSettings dsettings = (CalibrationSettings)d.getSettings();
CalibrationSettings dpsettings = (CalibrationSettings)dp.getSettings();
CvMat[] points1 = getPoints(useCenters);
CvMat[] points2 = peer.getPoints(useCenters);
// find points in common from points1 and points2
// since points1[0] and points2[0] might not be equal...
FloatBuffer objPts1 = points1[0].getFloatBuffer();
FloatBuffer imgPts1 = points1[1].getFloatBuffer();
IntBuffer imgCount1 = points1[2].getIntBuffer();
FloatBuffer objPts2 = points2[0].getFloatBuffer();
FloatBuffer imgPts2 = points2[1].getFloatBuffer();
IntBuffer imgCount2 = points2[2].getIntBuffer();
assert(imgCount1.capacity() == imgCount2.capacity());
CvMat objectPointsMat = CvMat.create(1, Math.min(objPts1.capacity(), objPts2.capacity()), CV_32F, 3);
CvMat imagePoints1Mat = CvMat.create(1, Math.min(imgPts1.capacity(), imgPts2.capacity()), CV_32F, 2);
CvMat imagePoints2Mat = CvMat.create(1, Math.min(imgPts1.capacity(), imgPts2.capacity()), CV_32F, 2);
CvMat pointCountsMat = CvMat.create(1, imgCount1.capacity(), CV_32S, 1);
FloatBuffer objectPoints = objectPointsMat.getFloatBuffer();
FloatBuffer imagePoints1 = imagePoints1Mat.getFloatBuffer();
FloatBuffer imagePoints2 = imagePoints2Mat.getFloatBuffer();
IntBuffer pointCounts = pointCountsMat .getIntBuffer();
int end1 = 0, end2 = 0;
for (int i = 0; i < imgCount1.capacity(); i++) {
int start1 = end1;
int start2 = end2;
end1 = start1 + imgCount1.get(i);
end2 = start2 + imgCount2.get(i);
int count = 0;
for (int j = start1; j < end1; j++) {
float x1 = objPts1.get(j*3 );
float y1 = objPts1.get(j*3+1);
float z1 = objPts1.get(j*3+2);
for (int k = start2; k < end2; k++) {
float x2 = objPts2.get(k*3 );
float y2 = objPts2.get(k*3+1);
float z2 = objPts2.get(k*3+2);
if (x1 == x2 && y1 == y2 && z1 == z2) {
objectPoints.put(x1);
objectPoints.put(y1);
objectPoints.put(z1);
imagePoints1.put(imgPts1.get(j*2));
imagePoints1.put(imgPts1.get(j*2+1));
imagePoints2.put(imgPts2.get(k*2));
imagePoints2.put(imgPts2.get(k*2+1));
count++;
break;
}
}
}
if (count > 0) {
pointCounts.put(count);
}
}
objectPointsMat.cols(objectPoints.position()/3);
imagePoints1Mat.cols(imagePoints1.position()/2);
imagePoints2Mat.cols(imagePoints2.position()/2);
pointCountsMat .cols(pointCounts .position());
// place our ProjectiveDevice at the origin...
d.R = CvMat.create(3, 3); d.R.put(1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 1.0);
d.T = CvMat.create(3, 1); d.T.put(0.0, 0.0, 0.0);
d.E = CvMat.create(3, 3); cvSetZero(d.E);
d.F = CvMat.create(3, 3); cvSetZero(d.F);
dp.R = CvMat.create(3, 3);
dp.T = CvMat.create(3, 1);
dp.E = CvMat.create(3, 3);
dp.F = CvMat.create(3, 3);
stereoCalibrate(getObjectPoints(objectPointsMat, pointCountsMat), getImagePoints(imagePoints1Mat, pointCountsMat), getImagePoints(imagePoints2Mat, pointCountsMat),
cvarrToMat(d.cameraMatrix), cvarrToMat(d.distortionCoeffs), cvarrToMat(dp.cameraMatrix), cvarrToMat(dp.distortionCoeffs),
new Size(d.imageWidth, d.imageHeight), cvarrToMat(dp.R), cvarrToMat(dp.T), cvarrToMat(dp.E), cvarrToMat(dp.F), dpsettings.flags,
new TermCriteria(CV_TERMCRIT_ITER+CV_TERMCRIT_EPS,100,1e-6));
// compute and return epipolar error...
d.avgEpipolarErr = 0.0;
d.maxEpipolarErr = 0.0;
double err[] = computeStereoError(imagePoints1Mat, imagePoints2Mat,
d.cameraMatrix, d.distortionCoeffs,
dp.cameraMatrix, dp.distortionCoeffs, dp.F);
dp.avgEpipolarErr = err[0];
dp.maxEpipolarErr = err[1];
return err;
}
}
© 2015 - 2025 Weber Informatics LLC | Privacy Policy