L6: 3D sensors

Question 1

What type of 3D acquisition methods is there?

Answer 1

Passive and active.

Question 2

What is passive?

Answer 2

Looks at the scene without doing something to is.

Question 3

What is active?

Answer 3

Sends light out, actively does something.
- A problem with active reflection methods is that the surface is non-reflective, meaning they won’t work.

Question 4

❗️❗️❗️What properties/advantages do active triangulation-based sensors have?

Answer 4

Pros:
- Passive/Non-contact, we are just observing the environment’s response to our stimulus (light)
- Fast
- Can estimate every point in the image

Cons:
- Susceptible to optical effect
- Mirrors/reflection
- High absorption (think of black Lego brick)
- Light penetrates a translucent object and scatter the rest of it

Question 5

❗️❗️❗️Explain Time-of-flight principle and phase difference method

Answer 5

Measures time it takes for an emitted signal to travel from the light sensor/source to a surface and back detected by the light emitter.
Range imaging camera system.

Phase difference: Does not measure distance directly, compares phase shift of two signals to determine the distance between them.
- Phase difference between the sent and received light is calculated

Question 6

❗️❗️❗️What are rectifying homographies and how can they be computed?

Answer 6

Determine two homographies one for each camera, so the rectification has the same row coordinate (scan lines) which can be directly matches.
- EPipolar lines are used as it intersect the epipolar plane that contains both camera centers and a point in the scene with the image plane.
- Reduces the search space.

Question 7

What is structued light?

Answer 7

Active lighting, projects one or more patterns, and one or more cameras, which capture images of these patterns projected onto the scene.

Question 8

What is scene coding?

Answer 8

Process of capturing and encoding information about the 3D structure and characteristics of a scene or environment

Question 9

What is stereo vision?

Answer 9

Passive
Uses a setup of two cameras to estimate the 3D point in space using triangulation.

Question 10

What is real time structured light?

Question 11

What are the three fundamental problems that occur during stereo matching?

Answer 11

• Occluding edges
• Homogeneous Areas
• Repetitive structure

Question 12

❗️❗️❗️Which directions need to be coded?

Answer 12

shoot out the planes vertical (column-wise) and if we shoot out more at once we need to encode them so we know which plane is which.

Question 13

❗️❗️❗️How does active scene coding provide correspondences?

Answer 13

We start by shooting rays, but does not give depth information and takes too long (many images too), so we projects lighting planes and multiple as it still taikes time. But there is ambiguity in depth so we use binary gray code.

Question 14

❗️❗️❗️Mention a few optical 3d sensing modalities.

Answer 14

• Multi-Pattern Structured light
• One-Shot Structured light
• Time of flight
• Stereo vision

The important part is that stereo is passive and the three others are active reflective methods.

Question 15

❗️❗️❗️Explain binary Gray coding and the phase shifting method.

Answer 15

Binary gray coding:
- Error: light bleeding is a thing = gray coding solves this by swapping by reducing the number of transitions
- Takes n images, each representing a bit for each stripe.
- All the stripes with bit 1 is on, with 0 off.
- n images, 2^n -1 stripes

Phase shifting method:
- project sinusoidal patterns and takes images to determine point, measures phase shift through the intensity of each point in each image.
- - Take the entinties to find the projected point, then project it as a plane afterwards.
- Uses fewer images (3)
- requires good calibration!