wgmmreg {LOMAR}

wgmmreg

Description

Rigid registration of two point sets by minimizing the Wasserstein distance between GMMs

Usage

wgmmreg(
  X,
  Y,
  CX,
  CY,
  wx = NULL,
  wy = NULL,
  maxIter = 200,
  subsample = NULL,
  tol = 1e-08
)

Arguments

X

reference point set, a N x D matrix

Y

point set to transform, a M x D matrix,

CX

array of covariance matrices for each point in X

CY

array of covariance matrices for each point in Y

wx

(optional) vector of mixture weights for X.

wy

(optional) vector of mixture weights for Y.

maxIter

maximum number of iterations to perform (default: 200)

subsample

if set, use this randomly selected fraction of the points

tol

tolerance for determining convergence (default: 1e-8)

Value

a list of

• Y: transformed point set,

• R: rotation matrix,

• t: translation vector,

• c: final value of the cost function,

• converged: logical, whether the algorithm converged.

Examples

data.file1 <- system.file("test_data", "parasaurolophusA.txt", package = "LOMAR",
 mustWork = TRUE)
PS1 <- read.csv(data.file1, sep = '\t', header = FALSE)
data.file2 <- system.file("test_data", "parasaurolophusB.txt", package = "LOMAR",
 mustWork = TRUE)
C1 <- diag(0.1, ncol(PS1)) + jitter(0.01, amount = 0.01)
C1 <- replicate(nrow(PS1),C1)
PS2 <- read.csv(data.file2, sep = '\t', header = FALSE)
C2 <- diag(0.1, ncol(PS2)) + jitter(0.01, amount = 0.01)
C2 <- replicate(nrow(PS2),C2)
transformation <- wgmmreg(PS1, PS2, C1, C2, subsample = 0.1, maxIter = 30, tol = 1e-4)
## Not run: 
# Visualize registration outcome
library(rgl)
plot3d(PS1, col = "blue")
points3d(PS2, col = "green")
points3d(transformation[['Y']], col = "magenta")

## End(Not run)