Skip to main content
. 2015 Dec 6;12(113):20150899. doi: 10.1098/rsif.2015.0899

Figure 1.

Figure 1.

Constructing eigenworms. In each video frame, thresholding is used to separate the animal from the background, then the resulting binary images are skeletonized. This skeleton, or midline, is used as a proxy for the animal's posture. Panel (a) shows a frame from the CBD with the worm's contour and midline highlighted, panel (b) shows the corresponding midline. The skeleton has been rotated to remove the worm's overall rotation relative to the plate. Panels (c) zooms in on the midline, showing how a set of θi angles provide a piecewise linear approximation to the midline curvature. This angular data forms a vector for each frame, or a matrix for a movie. The matrix's principal components are the eigenworms. Panel (d) shows an example of a posture reconstruction. The blue shapes in the middle column are the eigenworms, which can be added together with different weights to reconstruct any actual worm posture.