• Part 1: Estimate Homography
• Part 2: Unwarp the QR code on the screen
• Part 3: Estimate the 3D illusion
• Bonus: Simple Augmented Reality

Simply execute main.py to generate image for each part.

     python3 main.py

• Simple apply Homography Matrix for projection
• Reference (Wikipedia): Homgraphy

img1	img2	img3	img4	img5

Source	Result

• Apply Backward Warping to get pixel information (by interpolation)

Source	Result

• Apply Backward Warping to get the top view of the cross walk (pixel-wise interpolation is concerned)
• Discussion concerning the warping result is in the report.

Source	Result

• Apply Feature Detection and Projective Geometry

• Using SIFT as feature points detector

• Calculate Homography Matrix

• Algorithm Flow

  

• Usage

      python3 simple_ar.py -v [video]
                           -m [marker]
                           -i [input image]
                           -o [output video]
                           -d [for detect marker position]
  

• Example Usage

  • Output an projected image in the video (Default Parameters)

        python3 simple_ar.py -v ./input/ar_marker.mp4
                             -m ./input/marker.png
                             -i ./input/kp.jpg
                             -o output_kp.mp4
    

    Example output: output_kp.mp4

  • Output the detected position in the video

        python3 simple_ar.py -v ./input/ar_marker.mp4
                             -m ./input/marker.png
                             -o detectlines.mp4
                             -d
    

    Example output: detectlines.mp4