Measure a Planar Region with a Homography
Parking spaces on a flat pavement have different image shapes.
Measure a Planar Region with a Homography explained visually
Parking spaces on a flat pavement have different image shapes. English captions are optional and start off. The complete transcript follows the written lesson.
What you will learn
Parking spaces on a flat pavement have different image shapes.
A practical question
The parking camera sees a painted space as a slanted quadrilateral. On the ground, the marked rectangle has a known size. If its corners lie on the same plane, a homography can map between positions on that plane and positions in the image.
The useful idea
Match at least four suitable non-collinear point pairs, then estimate the projective mapping. Transform the space corners and inspect the overlay. The mapping can make a top-down view of the ground plane, but it cannot correctly flatten a tall car as though its roof were on the asphalt.
Work through the example
Measure a held-out painted segment in the rectified ground view and compare with its known length. Be careful near a poorly covered edge, where small corner mistakes can produce large measurement errors.
Limits and next step
Keep the source image, point pairs, and plane definition. Next, we'll search for distinctive image locations that we can match across changing views.
Worked example
Use four or more well-spread point correspondences on the ground plane.
Quick check: test your understanding
Try it yourself: In the situation described above, explain the method and one reason its output might be misleading.
Check the reasoning
A homography maps points on one plane between views, enabling a rectified pavement view. A tall moving car is off that plane and will not obey the ground homography.
What is the main idea behind measure a planar region with a homography?
A homography maps points on one plane between views, enabling a rectified pavement view.
How is it used in the example?
Use four or more well-spread point correspondences on the ground plane.
What limitation should be checked?
A tall moving car is off that plane and will not obey the ground homography.
Further reading
Programming Computer Vision with Python — Jan Erik Solem.
Original ToolsInVision teaching examples. Editorial approach · Suggest a correction
Accessible referenceLesson 24 transcript146 words · Video transcript · Matches the published narration and captionsOpen transcript
The parking camera sees a painted space as a slanted quadrilateral. On the ground, the marked rectangle has a known size. If its corners lie on the same plane, a homography can map between positions on that plane and positions in the image.
Match at least four suitable non-collinear point pairs, then estimate the projective mapping. Transform the space corners and inspect the overlay. The mapping can make a top-down view of the ground plane, but it cannot correctly flatten a tall car as though its roof were on the asphalt.
Measure a held-out painted segment in the rectified ground view and compare with its known length. Be careful near a poorly covered edge, where small corner mistakes can produce large measurement errors.
Keep the source image, point pairs, and plane definition. Next, we'll search for distinctive image locations that we can match across changing views.