Regresar

Evaluation of a protocol for automated extraction of morphometric measurements from avian eggs using digital photography

Abstract:

Evaluation of a protocol for automated extraction of morphometric measurements from avian eggs using digital photography. As many ecomorphological studies are limited by the time required to gather manual measurement data, automatizing the process is an important focus of methodological innovations. We developed, implemented and validated a protocol for the semi–automated extraction of a set of morphometric variables of egg size and shape from digital pictures. The protocol was implemented in R language as a web app called OvometriK. After binarizing and calibrating images, this protocol uses geometric and trigonometric functions to calculate eleven egg variables. We tested calculations in several ways, assuming contour continuity or using voxel counts. Application was validated with geometric shapes and 30 manually–measured chicken eggs. Mathematical validation with spheres showed that the algorithm provided high precision diameter measures, with a correlation of 99.9 %. Average estimation error was 1.4 %. The mathematical volume estimation was underestimated by 27 %, while voxels were underestimated by only 6 %. Differences between manual egg measurements of diameters and those obtained from images was less than 3 mm (4 %). Correlation between estimated volume and measured by silica gel filling was higher than 90 % using the voxel count method. Neither inclination angle or picture resolution had significant effects on precision (3.2 % maximum difference). Measures showed high repeatability and represent a significant saving in processing time. This new protocol represents an improvement on previous programs regarding limitations of platform, accessibility and number of variables. Furthermore, its flexibility and openness means it can be adapted to other specific applications.