imageviewset

Manage data for structure-from-motion, visual odometry, and visual SLAM

Since R2020a

Description

The imageviewset object manages view attributes and pairwise connections between views of data used in structure-from-motion, visual odometry, and simultaneous localization and mapping (SLAM) data. View attributes can be feature descriptors, feature points, or absolute camera poses. Pairwise connections between views can be point matches, relative camera poses, or an information matrix. You can also use this object to find point tracks used by triangulateMultiview and bundleAdjustment functions.

Creation

Syntax

vSet = imageviewset()

Description

vSet = imageviewset() returns an imageviewset object. You can add views and connections using the addView and addConnection object functions.

Properties

Views — View attributes

empty three-column table (default) | three-column table

This property is read-only.

View attributes, specified as a three-column table. The table contains columns as described in this table.

Column	Description
ViewId	View identifier for the view, specified as a unique integer
AbsolutePose	Absolute pose of the view, specified as a rigidtform3d object.
Features	Feature vectors, specified as an M-by-N matrix of M feature vectors or as a binaryFeatures.
Points	Image points, specified as one of these options: M-by-2 matrix of coordinates in the format [x,y] M-element feature point array. For more details, see Point Feature Types.

Connections — Pairwise connections between views

empty four-column table (default) | four-column table

This property is read-only.

Pairwise connections between views, specified as a four-column table. The table contains columns as described in this table. Each row corresponds to one connection.

Column	Description
ViewId1	View identifier for the first view, specified as a unique integer.
ViewId2	View identifier for the second view, specified as a unique integer.
RelativePose	Relative pose of the second view with respect to the first view, specified as a rigidtform3d or simtform3d object.
InformationMatrix	Information matrix, specified as a 6-by-6 matrix (when you specify RelativePose as a rigidtform3d object) or 7-by-7 matrix (when you specify RelativePose as a simtform3d object). The information matrix represents the uncertainty of the measurement error and is the inverse of the covariance matrix.
Matches	Indices of matched feature points between two views, specified as M-by-2 matrix.

NumViews — Number of views

0 (default) | nonnegative integer

This property is read-only.

Number of views, specified as a nonnegative integer.

NumConnections — Number of connections

0 (default) | nonnegative integer

This property is read-only.

Number of connections, specified as a nonnegative integer.

Object Functions

addView	Add views to view set
updateView	Update view in view set
deleteView	Delete view from view set
hasView	Check if view is in view set
addConnection	Add connection between views in view set
updateConnection	Update connection between views in a view set
deleteConnection	Delete a connection between views in view set
hasConnection	Check if connection between two views is in view set
findView	Find views associated with view identifiers
findConnection	Find connections associated with view identifiers
connectedViews	Returns directly and indirectly connected views
poses	Absolute poses associated with views in view set
createPoseGraph	Create pose graph
findTracks	Find matched points across multiple views
optimizePoses	Optimize absolute poses using relative pose constraints
plot	Plot image view set views and connections

Examples

Find Point Tracks Across Sequence of Images

Load images in the workspace.

imageDir = fullfile(toolboxdir('vision'),'visiondata','structureFromMotion');
images = imageDatastore(imageDir);

Compute features for the first image.

I = im2gray(readimage(images,1));
pointsPrev = detectSURFFeatures(I);
[featuresPrev,pointsPrev] = extractFeatures(I,pointsPrev);

Create an image view set and add one view to the set.

vSet = imageviewset;
vSet = addView(vSet,1,'Features',featuresPrev,'Points',pointsPrev);

Compute features and matches for the rest of the images.

for i = 2:numel(images.Files)
  I = im2gray(readimage(images, i));
  points = detectSURFFeatures(I);
  [features, points] = extractFeatures(I,points);
  vSet = addView(vSet,i,'Features',features,'Points',points);
  pairsIdx = matchFeatures(featuresPrev,features);
  vSet = addConnection(vSet,i-1,i,'Matches',pairsIdx);
  featuresPrev = features;
end

Find point tracks across views in the image view set.

tracks = findTracks(vSet);

Extended Capabilities

C/C++ Code Generation

Generate C and C++ code using MATLAB® Coder™.

The plot method of imageviewset object is not supported for code generation.

Version History

Introduced in R2020a

R2022b: Supports premultiply geometric transformation convention

Starting in R2022b, most Computer Vision Toolbox™ functions create and perform geometric transformations using the premultiply convention. Accordingly, the imageviewset object and its object functions are updated to use rigidtform3d and simtform3d objects, which support the premultiply convention. The imageviewset object now:

• Returns the Views property with the AbsolutePose column as a rigidtform3d object

• Returns the Connections property with the RelativePose column as a rigidtform3d or simtform3d object

Before, the properties returned AbsolutePose and RelativePose as rigid3d and affine3d objects, which use the postmultiply convention.

For more information, see Migrate Geometric Transformations to Premultiply Convention.

R2022b: Visual SLAM: Improve accuracy of pose estimation

The imageviewset adds the connectedViews object function.