Skip to main content

Table 3 A "one user" procedure of metric identification

From: The exchangeability of shape

 

Images (provided by multiple users)

Step 1

Obtaining reference images from a web data base

Step 2

Obtaining images of unknown specimens

 

Digitization (performed by a single user)

Step 3

Digitizing the images of reference (reference coordinates)

Step 4

Digitizing the images of the unknown specimens (unknown coordinates)

 

Classification

Procrustes

 

Step 5

Pairwise Procrustes distances between each unknown and each reference image

Mahalanobis

 

Step 6*

Computing shape variables on the combined sets of coordinates obtained from step 3 and one unknown specimen obtained from step 4

Step 7

Computing a discriminant model using the reference shape variables, exclusively (a partition of data from step 6)

Step 8

Entering to the discriminant model the shape variables of the unknown specimen (a partition of data from step 6)

Go to Step 6 for the next unknown specimen.

  1. General steps implemented in the CLIC package, relevant to the geometric approach: digitization (Steps 3 and 4, module COO of the CLIC package), the Procrustes classification (Step 5, module MOG of CLIC) and the discriminant analysis (Steps 6 and 7, module MOG) to identify organisms using mean reference pictures (Step 1) and own pictures (Step 2). The step 2 refers to the field and/or laboratory activities of the biologist: for entomologists, it generally requires the traditional tasks of collecting, dissecting and mounting insects. (*) To reduce multidimensionality, the number of shape variables (Step 6) can be reduced by selecting a set of few first relative warps (RW, i.e. principal components of partial warps). The MOG module automatically selects a number of RW lower than the smallest sample group.
Back to article page

BMC Research Notes

ISSN: 1756-0500

Contact us