Skip to content

Latest commit

 

History

History
53 lines (41 loc) · 2.47 KB

assignment.md

File metadata and controls

53 lines (41 loc) · 2.47 KB

Assignment: Camera Calibration

In this assignment, you will capture a set of images of a checkerboard pattern, use them to calibrate your camera, and then validate the calibration by undistorting a test image.

1. Prepare the Dataset

  • Capture 30–40 images of a checkerboard to use for calibration.
  • The checkerboard must be perfectly flat:
    • Display it on a flatscreen monitor or
    • Print it on paper and make sure it is completely unrolled and fixed to a flat surface.
  • Ensure the images are sufficiently diverse:
    • Vary the distance of the checkerboard from the camera.
    • Change angles and tilts.
    • Position the checkerboard near different parts of the image frame, especially the edges and corners.
  • Avoid blurry or over/under-exposed images.

2. Compute the Camera Model

  • Use OpenCV calibration functions (cv.findChessboardCorners, cv.calibrateCamera) to estimate the camera matrix and distortion coefficients.
  • Save the resulting parameters for later use.

3. Test the Calibration

  • Take one or more test images with the same camera and exact same settings as your calibration dataset:
    • Same resolution and aspect ratio.
    • Same focus and zoom level.
  • Undistort the test image using cv.undistort and compare the results.
  • ⚠️ Important note:
    • If your calibration dataset was created from video frames, you must also validate using frames from the same type of video, not separate still images.

Expected Submission

Your final report should include:

  1. Dataset and Validation Image Overview

    • Example images from the calibration dataset.
    • The test image used for validation.

  2. Camera Calibration

    • An image with detected checkerboard corners drawn on it.
    • The final camera model parameters:
      • Camera matrix
      • Distortion coefficients

  3. Testing

    • Side-by-side visualization of the undistorted image, showing:
      • Uncropped version (with black borders).
      • Cropped version (region of interest applied).

Hints

  • Use the example workflow in the notebook as a reference.
  • Start by testing with a small number of images to verify your pipeline.
  • Good calibration requires clear and accurate checkerboard corner detections.
  • For wide-angle cameras, consider using OpenCV’s fisheye calibration module for better results.