Difference between revisions of "Example Calibrate Planar Fisheye"

From BoofCV
Jump to navigationJump to search
m
m
(6 intermediate revisions by the same user not shown)
Line 7: Line 7:
This example demonstrates how to compute the intrinsic camera parameters for a fisheye camera lens. Fisheye lenses exhibit significantly more distortion than regular lenses with a more narrow field of view. Its not unusual for a fisheye lens to have a FOV of 185 degrees. The calibration process is very similar to regular cameras. A planar calibration target is shown at different angles across the entire field of view. The main difference is the camera model.
This example demonstrates how to compute the intrinsic camera parameters for a fisheye camera lens. Fisheye lenses exhibit significantly more distortion than regular lenses with a more narrow field of view. Its not unusual for a fisheye lens to have a FOV of 185 degrees. The calibration process is very similar to regular cameras. A planar calibration target is shown at different angles across the entire field of view. The main difference is the camera model.


Example File: [https://github.com/lessthanoptimal/BoofCV/blob/v0.28/examples/src/main/java/boofcv/examples/calibration/ExampleCalibrateFisheye.java ExampleCalibrateFisheye.java]
Example File: [https://github.com/lessthanoptimal/BoofCV/blob/v0.41/examples/src/main/java/boofcv/examples/calibration/ExampleCalibrateFisheye.java ExampleCalibrateFisheye.java]


Calibration Tutorial: [[Tutorial_Camera_Calibration|Wikipage]]
Calibration Tutorial: [[Tutorial_Camera_Calibration|Wikipage]]
Line 30: Line 30:
  * see at a distance and cover more of the fisheye's camera large FOV.
  * see at a distance and cover more of the fisheye's camera large FOV.
  *
  *
* @author Peter Abeles
  * @see CalibrateMonoPlanar
  * @see CalibrateMonoPlanar
*
* @author Peter Abeles
  */
  */
public class ExampleCalibrateFisheye {
public class ExampleCalibrateFisheye {
public static void main( String args[] ) {
public static void main( String[] args ) {
DetectorFiducialCalibration detector;
DetectSingleFiducialCalibration detector;
List<String> images;


// Circle based calibration targets not not recommended because the sever lens distortion will change
// Circle based calibration targets are not recommended because the sever lens distortion will change
// the apparent location of tangent points.
// the apparent location of tangent points.


// Square Grid example
// Square Grid example
// detector = FactoryFiducialCalibration.squareGrid(new ConfigSquareGrid(4, 3, 30, 30));
// detector = FactoryFiducialCalibration.squareGrid(null, new ConfigGridDimen(/*rows*/ 4, /*cols*/ 3, /*size*/ 30, /*space*/ 30));
// images = UtilIO.listAll(UtilIO.pathExample("calibration/fisheye/square_grid"));
// images = UtilIO.listAll(UtilIO.pathExample("calibration/fisheye/square_grid"));


// Chessboard Example
// Chessboard Example
detector = FactoryFiducialCalibration.chessboard(new ConfigChessboard(7, 5, 30));
detector = FactoryFiducialCalibration.chessboardX(null, new ConfigGridDimen(/*rows*/7, /*cols*/5, /*size*/30));
images = UtilIO.listAll(UtilIO.pathExample("calibration/fisheye/chessboard"));
List<String> images = UtilIO.listAll(UtilIO.pathExample("calibration/fisheye/chessboard"));


// Declare and setup the calibration algorithm
// Declare and setup the calibration algorithm
CalibrateMonoPlanar calibrationAlg = new CalibrateMonoPlanar(detector);
var calibrator = new CalibrateMonoPlanar();
 
// tell it type type of target and which parameters to estimate
calibrationAlg.configureUniversalOmni( true, 2, false);


// Specify the camera model to use. Here are a few examples.
//
calibrator.configureUniversalOmni( /*zeroSkew*/ true, /*radial*/ 2, /*tangential*/ false);
// it's also possible to fix the mirror offset parameter
// it's also possible to fix the mirror offset parameter
// 0 = pinhole camera. 1 = fisheye
// 0 = pinhole camera. 1 = fisheye
// calibrationAlg.configureUniversalOmni( true, 2, false,1.0);
// calibrationAlg.configureUniversalOmni( /*zeroSkew*/ true, /*radial*/ 2, /*tangential*/ false, /*offset*/ 1.0);
// Another popular model is Kannala-Brandt. Most people just use the symmetric terms.
// calibrationAlg.configureKannalaBrandt( /*zeroSkew*/ true, /*symmetric*/ 5, /*asymmetric*/ 0);


for( String n : images ) {
var usedImages = new ArrayList<String>();
for (String n : images) {
BufferedImage input = UtilImageIO.loadImage(n);
BufferedImage input = UtilImageIO.loadImage(n);
if( input != null ) {
if (input == null)
GrayF32 image = ConvertBufferedImage.convertFrom(input,(GrayF32)null);
continue;
if( !calibrationAlg.addImage(image) )
GrayF32 image = ConvertBufferedImage.convertFrom(input, (GrayF32)null);
System.err.println("Failed to detect target in "+n);
if (detector.process(image)) {
// Need to tell it the image shape and the layout once
if (usedImages.isEmpty())
calibrator.initialize(image.getWidth(), image.getHeight(), detector.getLayout());
calibrator.addImage(detector.getDetectedPoints().copy());
usedImages.add(n);
} else {
System.err.println("Failed to detect target in " + n);
}
}
}
}
// process and compute intrinsic parameters
// process and compute intrinsic parameters
CameraUniversalOmni intrinsic = calibrationAlg.process();
CameraModel intrinsic = calibrator.process();


// save results to a file and print out
// save results to a file and print out
CalibrationIO.save(intrinsic, "fisheye.yaml");
CalibrationIO.save(intrinsic, "fisheye.yaml");


calibrationAlg.printStatistics();
System.out.println(calibrator.computeQualityText(usedImages));
System.out.println();
System.out.println();
System.out.println("--- Intrinsic Parameters ---");
System.out.println("--- Intrinsic Parameters ---");
Line 81: Line 89:
}
}
}
}
</syntaxhighlight>
</syntaxhighlight>

Revision as of 13:19, 2 September 2022

This example demonstrates how to compute the intrinsic camera parameters for a fisheye camera lens. Fisheye lenses exhibit significantly more distortion than regular lenses with a more narrow field of view. Its not unusual for a fisheye lens to have a FOV of 185 degrees. The calibration process is very similar to regular cameras. A planar calibration target is shown at different angles across the entire field of view. The main difference is the camera model.

Example File: ExampleCalibrateFisheye.java

Calibration Tutorial: Wikipage

Concepts:

  • Camera calibration
  • Fisheye Lens distortion
  • Intrinsic parameters

Relevant Videos:

Related Examples:

Example Code

/**
 * Example of how to calibrate a single (monocular) fisheye camera using a high level interface. This example
 * for the most part follows the same routine as {@link ExampleCalibrateMonocular}. Fisheye cameras tend to require
 * more images to properly calibrate. Often people will use larger calibration targets too that are easier to
 * see at a distance and cover more of the fisheye's camera large FOV.
 *
 * @author Peter Abeles
 * @see CalibrateMonoPlanar
 */
public class ExampleCalibrateFisheye {
	public static void main( String[] args ) {
		DetectSingleFiducialCalibration detector;

		// Circle based calibration targets are not recommended because the sever lens distortion will change
		// the apparent location of tangent points.

		// Square Grid example
//		detector = FactoryFiducialCalibration.squareGrid(null, new ConfigGridDimen(/*rows*/ 4, /*cols*/ 3, /*size*/ 30, /*space*/ 30));
//		images = UtilIO.listAll(UtilIO.pathExample("calibration/fisheye/square_grid"));

//		 Chessboard Example
		detector = FactoryFiducialCalibration.chessboardX(null, new ConfigGridDimen(/*rows*/7, /*cols*/5, /*size*/30));
		List<String> images = UtilIO.listAll(UtilIO.pathExample("calibration/fisheye/chessboard"));

		// Declare and setup the calibration algorithm
		var calibrator = new CalibrateMonoPlanar();

		// Specify the camera model to use. Here are a few examples.
		//
		calibrator.configureUniversalOmni( /*zeroSkew*/ true, /*radial*/ 2, /*tangential*/ false);
		// it's also possible to fix the mirror offset parameter
		// 0 = pinhole camera. 1 = fisheye
//		calibrationAlg.configureUniversalOmni( /*zeroSkew*/ true, /*radial*/ 2, /*tangential*/ false, /*offset*/ 1.0);
		// Another popular model is Kannala-Brandt. Most people just use the symmetric terms.
//		calibrationAlg.configureKannalaBrandt( /*zeroSkew*/ true, /*symmetric*/ 5, /*asymmetric*/ 0);

		var usedImages = new ArrayList<String>();
		for (String n : images) {
			BufferedImage input = UtilImageIO.loadImage(n);
			if (input == null)
				continue;
			GrayF32 image = ConvertBufferedImage.convertFrom(input, (GrayF32)null);
			if (detector.process(image)) {
				// Need to tell it the image shape and the layout once
				if (usedImages.isEmpty())
					calibrator.initialize(image.getWidth(), image.getHeight(), detector.getLayout());
				calibrator.addImage(detector.getDetectedPoints().copy());
				usedImages.add(n);
			} else {
				System.err.println("Failed to detect target in " + n);
			}
		}
		// process and compute intrinsic parameters
		CameraModel intrinsic = calibrator.process();

		// save results to a file and print out
		CalibrationIO.save(intrinsic, "fisheye.yaml");

		System.out.println(calibrator.computeQualityText(usedImages));
		System.out.println();
		System.out.println("--- Intrinsic Parameters ---");
		System.out.println();
		intrinsic.print();
	}
}