estimateGeometricTransform3D

(Not recommended) Estimate 3-D geometric transformation from matching point pairs

estimateGeometricTransform3D is not recommended. Use the estgeotform3d function instead. For more information, see Version History.

Syntax

tform = estimateGeometricTransform3D(matchedPoints1,matchedPoints2,transformType)

[tform,inlierIndex] = estimateGeometricTransform3D(___)

[tform,inlierIndex,status] = estimateGeometricTransform3D(___)

[___] = estimateGeometricTransform3D(___, Name,Value)

Description

tform = estimateGeometricTransform3D(matchedPoints1,matchedPoints2,transformType) estimates a 3-D geometric transformation between two sets of 3-D points by mapping the inliers in the matched points from one set of 3-D points matchedPoints1 to the inliers in the matched points from the other set of 3-D points matchedPoints2.

[tform,inlierIndex] = estimateGeometricTransform3D(___) additionally returns a vector specifying each matched point pair as either an inlier or an outlier using the input arguments from the previous syntax.

[tform,inlierIndex,status] = estimateGeometricTransform3D(___) additionally returns a status code indicating whether or not the function could estimate a transformation and, if not, why it failed. If you do not specify the status output, the function instead returns an error for conditions that cannot produce results.

[___] = estimateGeometricTransform3D(___, Name,Value) specifies additional options using one or more name-value arguments in addition to any combination of arguments from previous syntaxes. For example, "Confidence",99 sets the confidence value for finding the maximum number of inliers to 99.

Examples

Estimate Transformation Between Two Point Clouds

Load a point cloud file into the workspace.

ptCloud1 = pcread('teapot.ply')

ptCloud1 = 
  pointCloud with properties:

     Location: [41472×3 single]
        Count: 41472
      XLimits: [-3 3.4340]
      YLimits: [-2 2]
      ZLimits: [0 3.1500]
        Color: []
       Normal: []
    Intensity: []

ptCloud1 = pcdownsample(ptCloud1,'random',0.25);  

Create a rigid 3-D transformation object with a 30-degree rotation.

theta = 30; % degrees
rot = [cosd(theta)  sind(theta) 0; ...
      -sind(theta)  cosd(theta) 0; ...
            0           0       1];
trans = [0 0 0];
tform = rigid3d(rot,trans);

Transform the point cloud using the transformation object.

ptCloud2 = pctransform(ptCloud1,tform);

To introduce noise, add random points to both point clouds.

noise1 = rescale(rand(1000,3),-2,2);
ptCloud1 = pointCloud([ptCloud1.Location;noise1]);
noise2 = rescale(rand(1000,3),-2,2);
ptCloud2 = pointCloud([ptCloud2.Location;noise2]);

Visualize the noisy point clouds.

figure
pcshowpair(ptCloud1,ptCloud2)
title('Point Clouds With Added Noise')

Extract matched points from the point clouds.

matchedPoints1 = ptCloud1.Location;
matchedPoints2 = ptCloud2.Location;

Estimate the rigid transformation between the point clouds.

[tformEst,inlierIndex] = estimateGeometricTransform3D(matchedPoints1, ...
    matchedPoints2,'rigid'); 

Extract the inlier points.

inliersPtCloud1 = transformPointsForward(tformEst,matchedPoints1(inlierIndex,:));
inliersPtCloud2 = matchedPoints2(inlierIndex,:);

Visualize the inliers of the aligned point clouds.

figure
firstPtCloud = pointCloud(inliersPtCloud1);
secondPtCloud = pointCloud(inliersPtCloud2);
pcshowpair(firstPtCloud,secondPtCloud)
title('Aligned point clouds')

Input Arguments

matchedPoints1 — First set of matched 3-D points

M-by-3 matrix of [x,y,z] coordinates

First set of matched 3-D points, specified as an M-by-3 matrix in which each row is a set of (x,y,z) coordinates and M is the number of matched points.

matchedPoints2 — Second set of matched 3-D points

M-by-3 matrix of [x,y,z] coordinates

Second set of matched 3-D points, specified as an M-by-3 matrix in which each row is a set of (x,y,z) coordinates and M is the number of matched points.

transformType — Transformation type

"rigid" | "similarity"

Transformation type, specified as "rigid" or "similarity". Each transform type requires a minimum number of matched pairs of points to estimate a transformation. You can generally improve the accuracy of a transformation by using a larger number of matched pairs of points. This table shows the type of object associated with each transformation type and the minimum number of matched pairs of points the transformation requires.

transformType	tform Object	Minimum Number of Matched Pairs of Points
"rigid"	rigid3d	3
"similarity"	affine3d	3

Data Types: string

Name-Value Arguments

Specify optional pairs of arguments as Name1=Value1,...,NameN=ValueN, where Name is the argument name and Value is the corresponding value. Name-value arguments must appear after other arguments, but the order of the pairs does not matter.

Before R2021a, use commas to separate each name and value, and enclose Name in quotes.

Example: "Confidence",99 sets the confidence value for finding the maximum number of inliers to 99.

MaxNumTrials — Maximum random trials

1000 (default) | positive integer

Maximum number of random trials, specified as a positive integer. This value specifies the number of randomized attempts the function makes to find matching point pairs. Specifying a higher value causes the function to perform additional computations, which increases the likelihood of finding inliers.

Data Types: single | double | int8 | int16 | int32 | int64 | uint8 | uint16 | uint32 | uint64

Confidence — Confidence of finding maximum number of inliers

99 (default) | positive numeric scalar

Confidence of finding the maximum number of inliers, specified as a positive numeric scalar in the range (0, 100). Increasing this value causes the function to perform additional computations, which increases the likelihood of finding a greater number of inliers.

Data Types: single | double | int8 | int16 | int32 | int64 | uint8 | uint16 | uint32 | uint64

MaxDistance — Maximum distance from point to projection

1.5 (default) | positive numeric scalar

Maximum distance from a point to the projection of the corresponding point, specified as a positive numeric scalar. "MaxDistance" specifies the maximum distance, in pixels, that a point can differ from the projected location of its corresponding point to be considered an inlier. The corresponding projection is based on the estimated transform.

The function checks for a transformation from matchedPoints1 to matchedPoints2, and then calculates the distance between the matched points in each pair after applying the transformation. If the distance between the matched points in a pair is greater than the "MaxDistance" value, then the pair is considered an outlier for that transformation. If the distance is less than "MaxDistance", then the pair is considered an inlier.

Data Types: single | double | int8 | int16 | int32 | int64 | uint8 | uint16 | uint32 | uint64

Output Arguments

tform — Geometric transformation

rigid3d object | affine3d object

Geometric transformation, returned as an affine3d or a rigid3d object.

The returned geometric transformation matrix maps the inliers in matchedPoints1 to the inliers in matchedPoints2. The function returns an object specific to the transformation type specified by the transformType input argument.

transformType	tform
"rigid"	rigid3d
"similarity"	affine3d

inlierIndex — Inliers

M-by-1 logical vector

Inliers, returned as an M-by-1 logical vector of M point pairs. Each element contains either a logical true, 1, to indicate the point pair is an inlier, or a logical false, 0, to indicate the point pair is an outlier.

status — Status code

0 | 1 | 2

Status code, returned as 0, 1, or 2. The status code indicates whether or not the function could estimate the transformation and, if not, why it failed.

Value	Description
0	No error
1	matchedPoints1 and matchedPoints2 inputs do not contain enough points
2	Not enough inliers found

If you do not specify the status code output, the function returns an error if it cannot produce results.

Data Types: int32

Algorithms

The function excludes outliers using the M-estimator sample consensus (MSAC) algorithm. The MSAC algorithm is a variant of the random sample consensus (RANSAC) algorithm. Results may not be identical between runs due to the randomized nature of the MSAC algorithm.

References

[1] Hartley, Richard, and Andrew Zisserman. Multiple View Geometry in Computer Vision. 2nd ed. Cambridge, UK ; New York: Cambridge University Press, 2003.

[2] Torr, P.H.S., and A. Zisserman. "MLESAC: A New Robust Estimator with Application to Estimating Image Geometry." Computer Vision and Image Understanding. 78, no. 1 (April 2000): 138–56. https://doi.org/10.1006/cviu.1999.0832.

Extended Capabilities

C/C++ Code Generation
Generate C and C++ code using MATLAB® Coder™.

Version History

Introduced in R2020b

R2022b: Not recommended

Starting in R2022b, many Computer Vision Toolbox™ functions create and perform geometric transformations using the premultiply convention. However, the estimateGeometricTransform3d function estimates geometric transformations using the postmultiply convention. Although there are no plans to remove estimateGeometricTransform3d at this time, you can streamline your geometric transformation workflows by switching to the estgeotform3d function, which supports the premultiply convention. For more information, see Migrate Geometric Transformations to Premultiply Convention.

To update your code, change instances of the function name estimateGeometricTransform3d to estgeotform3d. You do not need to change the other arguments.