SfmTrack2d - GTSAM Docs

Created by Codex.

Represent one structure-from-motion (SfM) landmark track as camera-indexed two-dimensional observations.

Authors: Frank Dellaert, et al. (see THANKS for the full author list)

See LICENSE for the license information

Mathematical idea¶

A two-dimensional track is an indexed observation set

\mathcal T_j^{2D}=\{(i,z_{ij})\},\qquad z_{ij}\in\mathbb R^2,

(1)

where image index $i$ may occur at most once. Triangulation later associates these measurements with one three-dimensional landmark $P_j$ .

try:
    import google.colab
    %pip install --quiet gtsam-develop
except ImportError:
    pass

import gtsam
import numpy as np

from gtsam import symbol_shorthand

C = symbol_shorthand.C
K = symbol_shorthand.K
P = symbol_shorthand.P
S = symbol_shorthand.S
X = symbol_shorthand.X

Core operations¶

Each measurement is (camera_index, Point2). measurementMatrix() returns an N x 2 array and indexVector() returns the corresponding camera indices. siftIndex() uses the legacy Scale-Invariant Feature Transform (SIFT) name for the feature position within the track; it is not an external feature identifier.

track = gtsam.SfmTrack2d()
track.addMeasurement(0, np.array([320.0, 240.0]))
track.addMeasurement(2, np.array([301.5, 241.0]))

print("measurements:", track.numberMeasurements())
print("camera indices:", track.indexVector())
print("image points:")
print(track.measurementMatrix())
print("unique cameras:", track.hasUniqueCameras())

measurements: 2
camera indices: [0 2]
image points:
[[320.  240. ]
 [301.5 241. ]]
unique cameras: True

Practical notes¶

Camera indices refer to positions in an SfmData camera list, not arbitrary GTSAM keys. Use gtsam.gtsfm.tracksFromPairwiseMatches to assemble tracks from pairwise feature matches.

Source¶

SfmTrack.h