Scribd
Upload
291 views19 pages

Stereo Vision

Stereo vision uses two or more camera views of the same scene to extract depth information and generate 3D point clouds. It works by comparing the two images and estimating the relative depth of points. This process outputs a 3D point cloud where each point corresponds to image pixels and contains depth information. Stereo vision is used in applications like advanced driver assistance and robot navigation to estimate object distances from cameras. It provides depth perception similar to human stereo vision from two eyes having slightly different views. Accurate camera calibration is important for stereo vision systems to properly relate the 3D real world to 2D images.

Uploaded by

Ayesha Hussain
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
291 views19 pages

Stereo Vision

Stereo vision uses two or more camera views of the same scene to extract depth information and generate 3D point clouds. It works by comparing the two images and estimating the relative depth of points. This process outputs a 3D point cloud where each point corresponds to image pixels and contains depth information. Stereo vision is used in applications like advanced driver assistance and robot navigation to estimate object distances from cameras. It provides depth perception similar to human stereo vision from two eyes having slightly different views. Accurate camera calibration is important for stereo vision systems to properly relate the 3D real world to 2D images.

Uploaded by

Ayesha Hussain
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 19

Stereo Vision

Stereo Vision
• Stereo Vision is a process of comparing two or
more views of same images and recovering
the depth information from the camera
images by estimating the relative depth of
point in the scene.
Stereo Vision
• The output of this computation id 3D point
cloud where each point corresponds to the
pixels of one of the image.
Stereo Vision
• It is used in the application such as advanced
driver assistant systems ADA’s or robot
navigation where stereo vision is used to
estimate the actual distance or range of
objects of interest from the camera.
Stereo Vision
• Human have two eyes side by side with a little distance b/w
them. This provide each eye with a different angle of same view
area. We can verify this if we try holding our finger in front of our
face and try alternating opening and closing one eye at a time.

•
Stereo Vision
• The brain process the images from both the eye combining them
into a 3D image with depth information
• Stereo vision cameras use the same principle to estimate depth.

•
Stereo Vision
• If we take a look at this image as humans we
can immediately perceive the depth in it. And
see is closer to us than the man in the black
shirt and it could be great I we use this
information to extract object from an image.
Stereo Vision
• The good news is we can. Stereo vision allows
us to do that.
• It gives us the point cloud from the images
from which we can specify the depth
information to extract out just the lady and
filter out the rest of the image. Or may be w
can specify the depth information for the
gentleman and keep only him and his
surroundings.
Stereo Vision
Stereo Vision
• So now that we have good understanding of
what the stereo vision is, how it works and how
it can useful for our applications, we need to
know how to acquire stereo images to
• To acquire stereo images we can either use an
integrated stereo vision camera such as the Z or
the bumble b 2 or we can use multiple cameras
• #Example: Two USB cameras
Stereo Vision
Stereo Vision
• The advantage to using an integrate stereo vision camera is that it
offers less error than a custom multiple camera setup. An integrated
stereo vision camera provide hardware triggering such that two
images are taking from exactly same moment with multiple camera
setup images must be taken at a time from each camera.
• The lenses in an integrated stereo vision camera do not move with
respect to each other which eliminates motion artifacts that occur if
you use individual cameras and if an object moves between captures
• This means that the camera calibration always remains accurate.
• A rig needs to be designed to ensure that in a multiple camera setup
the camera do not move that a shift oscillate or shake with respect to
each other.
Stereo Vision
• From mono vision cameras camera calibration
is an important one.
Stereo Vision
• From mono vision cameras camera calibration is an important one.
• An image is a 2D representation of 3D world scene.
• The mapping from world to image can be described
mathematically by a series of transformations
• The goal of camera calibration is to estimate the parameters for
these transformations
• The extrinsic parameters describe the location of camera
coordinates systems within the world coordinate system.
• So basically where the camera is placed in the world.
• The intrisic parameters are required for mapping from 3D camera
coordinates to 2D image coordinates.
Stereo Vision
• For stereo vision cameras accurate camera
calibrations becomes even more essential
because we can then use it to recover depth
from images.
Stereo Vision
• A stereo system consists of two cameras. So in
addition to traditional camera calibration we
also need to calibrate the orientation of
camera two relative to camera one.
• In Mat lab there is a stereo camera calibrated
app which estimates the parameters of each
of the two cameras.
Stereo Vision
• There are six steps in the process of stereo
camera calibration
Stereo Vision
• The first step is prepare images needs to prepare a checkerboard
pattern. Now we add the to the stereo calibrator app. The app
then calibrate the stereo parameters for us. Now we can
evaluate the results by looking at the re-projection error which
are the distances and pixels between the detected and the re-
projected points.
• The app first detect the checkerboard pints from the image
pairs.it the calculate the re projection error by projecting the
checkerboard points from world coordinates defined by the
checkerboard into image coordinated .
• It then compare the reprojected points to the corresponding
detecting points
Stereo Vision
As a general rule re-projection error is less than 1 pixel are
acceptable.
We can also look at 3d extrinsic parameters plot which provide
camera centric view of the patterns to evaluate calibration
accuracy.
The camera centric view is helpful if the camera was stationary
when the images were captured.
This view help us to examine the relative positions of the patterns
and the camera to see if they match.
We expect we can remove outlier images and recalibrate to
improve results and finally export the calibrator stereo parameters

You might also like