ECE 595 Computer Vision Assignment 5

Submit Matlab code (published and original), images, and a Word document describing your work.

1. Follow the steps in Example 1 from the Camera Calibration Toolbox for MatlabŪ by Jean-Yves Bouguet.

2. The attachments include data for the robot position and orientation for the 2D planar translation and rotation studies from assignment 4. Correlate your results with this ground truth and plot your errors as a function of robot truth. See assignment 4 notes for a suggested format.

3. Correct the translation images for distortion and repeat the measurement of 2D planar translation studies. Ideally this will eliminate the errors due to distortion in your final results.

4. Use MATLAB to reproduce the photographs taken of various products on a checkerboard background. (see attachments) Construct a MATLAB 3D model of your object, from assignment 4, or if you did not have your object on display, use an object from a fellow student. Adjust the camera positions (trial and error methods are acceptable) to create an image as close as you can to your selected photograph. It is not necessary to re-submit the images for the six faces of the product.

5. Use the checked cloth image (see example below) as input and construct a view orthonormal to the checker pattern. The original images are in the attachments.

   

6. Use your cube model and set MATLAB parameters for a perspective view. Follow the notes in cube6 and modify the code as necessdary to reconstruct the known camera parameters.

7. Use the first image in the sequence of images shown below to calibrate an orthogonal view of the floor. Use that TFORM to transform the other images. Determine the distance the target moved from image to image. The original images are in the attachments.

Maintained by John Loomis, last updated 23 March 2011