dynamic perspective
play

Dynamic Perspective CS 543 / ECE 549 Saurabh Gupta Spring 2020, - PowerPoint PPT Presentation

Dynamic Perspective CS 543 / ECE 549 Saurabh Gupta Spring 2020, UIUC http://saurabhg.web.illinois.edu/teaching/ece549/sp2020/ Many slides adapted from J. Malik. Perspective Projection P Y &' ( , &) , , ( Z f


  1. Dynamic Perspective CS 543 / ECE 549 – Saurabh Gupta Spring 2020, UIUC http://saurabhg.web.illinois.edu/teaching/ece549/sp2020/ Many slides adapted from J. Malik.

  2. Perspective Projection P Y &' ( , &) π‘Œ, 𝑍, π‘Ž β†’ ( Z f O y Suppose the camera moves with p respect to the world… β€’ Point P (π‘Œ, 𝑍, π‘Ž) in the world moves relative to the camera, its projection in the image (𝑦, 𝑧) moves as well. β€’ This movement in the image plane is called optical flow. β€’ Suppose the image of the point (𝑦, 𝑧) moves to (𝑦 + βˆ†π‘¦, 𝑧 + βˆ†π‘§) in time βˆ†π‘’ , then 12 13 , 14 are the two 13 components of the optical flow.

  3. Outline β€’ Relate optical flow to camera motion β€’ Special cases

  4. Μ‡ Μ‡ Μ‡ How does a point X in the scene move? β€’ Assume that the camera moves with a translational velocity 𝑒 = (𝑒 2 , 𝑒 4 , 𝑒 6 ) and angular velocity πœ• = πœ• 2 , πœ• 4 , πœ• 6 . β€’ Linear velocity of point 𝑄 = π‘Œ, 𝑍, π‘Ž is given 𝑄 = βˆ’π‘’ βˆ’ πœ•Γ—π‘„. by Μ‡ πœ• 4 π‘Ž βˆ’ πœ• 6 𝑍 𝑒 2 π‘Œ 𝑒 4 πœ• 6 π‘Œ βˆ’ πœ• 2 π‘Ž = βˆ’ βˆ’ 𝑍 𝑒 6 πœ• 2 𝑍 βˆ’ πœ• 4 π‘Œ π‘Ž

  5. Μ‡ Μ‡ Μ‡ Μ‡ Μ‡ Μ‡ Μ‡ Μ‡ Now, lets consider the effect of projection πœ• 4 π‘Ž βˆ’ πœ• 6 𝑍 𝑒 2 π‘Œ 𝑒 4 πœ• 6 π‘Œ βˆ’ πœ• 2 π‘Ž = βˆ’ βˆ’ 𝑍 𝑒 6 πœ• 2 𝑍 βˆ’ πœ• 4 π‘Œ π‘Ž ' ) β€’ Assume, 𝑔 = 1 , 𝑦 = ( , 𝑧 = ( . '(@ Μ‡ )(@ Μ‡ (' () β€’ 𝑦 = 𝑧 = , Μ‡ ( A ( A π‘Œ, Μ‡ 𝑍, Μ‡ π‘Ž , from equation above: β€’ Substitute Μ‡ πœ• 2 𝑒 2 βˆ’ 1 + 𝑦 E 𝑧 = 1 𝑦𝑧 𝑧 𝑣 𝑦 βˆ’1 0 𝑦 𝑒 4 πœ• 4 𝑀 = + 0 βˆ’1 𝑧 (1 + 𝑧 E ) π‘Ž βˆ’π‘¦π‘§ βˆ’π‘¦ πœ• 6 𝑒 6

  6. Μ‡ Μ‡ Dynamic Perspective Equations 𝑒 2 πœ• 2 βˆ’ 1 + 𝑦 E 𝑧 = 1 𝑦𝑧 𝑧 𝑣 𝑦 βˆ’1 0 𝑦 𝑒 4 πœ• 4 𝑀 = + 0 βˆ’1 𝑧 (1 + 𝑧 E ) π‘Ž βˆ’π‘¦π‘§ βˆ’π‘¦ πœ• 6 𝑒 6 Translation Component Rotation Component

  7. Μ‡ Μ‡ Optical flow for pure rotation 𝑒 2 πœ• 2 βˆ’ 1 + 𝑦 E 𝑧 = 1 𝑦𝑧 𝑧 𝑣 𝑦 βˆ’1 0 𝑦 𝑒 4 πœ• 4 𝑀 = + 0 βˆ’1 𝑧 (1 + 𝑧 E ) π‘Ž βˆ’π‘¦π‘§ βˆ’π‘¦ πœ• 6 𝑒 6 πœ• 2 βˆ’ 1 + 𝑦 E 𝑦𝑧 𝑧 𝑣 πœ• 4 𝑀 = β€’ (1 + 𝑧 E ) βˆ’π‘¦π‘§ βˆ’π‘¦ πœ• 6 β€’ We can determine πœ• from the flow field. β€’ Flow field is independent of π‘Ž(𝑦, 𝑧) .

  8. Μ‡ Μ‡ Optical flow for pure translation along Z-axis 𝑒 2 πœ• 2 βˆ’ 1 + 𝑦 E 𝑧 = 1 𝑦𝑧 𝑧 𝑣 𝑦 βˆ’1 0 𝑦 𝑒 4 πœ• 4 𝑀 = + 0 βˆ’1 𝑧 (1 + 𝑧 E ) π‘Ž βˆ’π‘¦π‘§ βˆ’π‘¦ πœ• 6 𝑒 6 𝑦 𝑣 3 F β€’ 𝑀 = 𝑧 ((2,4) β€’ Optical flow vector is a scalar multiple of position vector. β€’ Scale factor ambiguity, if 𝑒 6 β†’ 𝑙𝑒 6 , and π‘Ž β†’ π‘™π‘Ž , optical flow remains unchanged. β€’ But, you can get time to collision, π‘Ž/𝑒 6 .

  9. Μ‡ Μ‡ Optical flow for general translation 𝑒 2 πœ• 2 βˆ’ 1 + 𝑦 E 𝑧 = 1 𝑦𝑧 𝑧 𝑣 𝑦 βˆ’1 0 𝑦 𝑒 4 πœ• 4 𝑀 = + 0 βˆ’1 𝑧 (1 + 𝑧 E ) π‘Ž βˆ’π‘¦π‘§ βˆ’π‘¦ πœ• 6 𝑒 6 @3 L J43 F β€’ 𝑣 = @3 I J23 F ((2,4) , v = ((2,4)

  10. Optical flow for points on a road Slide by J. Malik

  11. Translating along X-axis in front of a wall Slide by J. Malik

  12. Estimating Optical Flow from Images http://en.wikipedia.org/wiki/Barberpole_illusion

  13. Estimating Optical Flow from Images http://en.wikipedia.org/wiki/Barberpole_illusion

  14. Estimating Optical Flow from Images Aperture Problem

  15. Μ‡ Μ‡ Recap β€’ Relate optical flow to camera motion 𝑒 2 πœ• 2 βˆ’ 1 + 𝑦 E 𝑧 = 1 𝑦𝑧 𝑧 𝑣 𝑦 βˆ’1 0 𝑦 𝑒 4 πœ• 4 𝑀 = + 0 βˆ’1 𝑧 (1 + 𝑧 E ) π‘Ž βˆ’π‘¦π‘§ βˆ’π‘¦ πœ• 6 𝑒 6 β€’ Special cases β€’ Pure rotation / pure translation / time to collision

Recommend


More recommend