Source code for attitude.geom

import numpy as N
from .util import dot
from .conics import conic
from .util import angle, vector
from .transform import rotate_2D

def aligned_covariance(fit, type='noise'):
    """
    Covariance rescaled so that eigenvectors sum to 1
    and rotated into data coordinates from PCA space
    """
    cov = fit._covariance_matrix(type)
    # Rescale eigenvectors to sum to 1
    cov /= N.linalg.norm(cov)
    return dot(fit.axes,cov)

def fit_angle(fit1, fit2, degrees=True):
    """
    Finds the angle between the nominal vectors
    """
    return N.degrees(angle(fit1.normal,fit2.normal))

def fit_similarity(fit1, fit2):
    """
    Distance apart for vectors, given in
    standard deviations
    """
    cov1 = aligned_covariance(fit1)
    cov2 = aligned_covariance(fit2)
    if fit2.axes[2,2] < 0:
        cov2 *= -1
    v0 = fit1.normal-fit2.normal
    cov0 = cov1+cov2 # Axes are aligned, so no covariances

    # Find distance of point from center

    # Decompose covariance matrix
    U,s,V = N.linalg.svd(cov0)
    rotated = dot(V,v0) # rotate vector into PCA space
    val = rotated**2/N.sqrt(s)
    return N.sqrt(val.sum())

def axis_aligned_transforms():
    """
    Generates three transformation matrices
    to map three-dimensional data down
    to slices on the xy, xz and yz planes, respectively.
    The typical use case for this is to iterate over the results
    of this method to provide arguments to e.g. the `HyperbolicErrors`
    function.
    """
    I = N.eye(3)
    xy = I[:2]
    xz = N.vstack((I[0],I[2]))
    yz = I[1:]
    return xy,xz,yz