Open AccessTowards the automated evaluation of crystallization trials
Author(s) -
Wilson Julie
Publication year - 2002
Publication title -
acta crystallographica section d
Language(s) - English
Resource type - Journals
ISSN - 1399-0047
DOI - 10.1107/s0907444902016633
Subject(s) - classification of discontinuities , boundary (topology) , object (grammar) , curvature , artificial intelligence , image (mathematics) , enhanced data rates for gsm evolution , mathematics , computer science , intensity (physics) , variance (accounting) , computer vision , pattern recognition (psychology) , algorithm , geometry , optics , mathematical analysis , physics , accounting , business
A method to evaluate images from crystallization experiments is described. Image discontinuities are used to determine boundaries of artifacts in the images and these are then considered as individual objects. This allows the edge of the drop to be identified and any objects outside this ignored. Each object is evaluated in terms of a number of attributes related to its size and shape, the curvature of the boundary and the variance in intensity, as well as obvious crystal‐like characteristics such as straight sections of the boundary and straight lines of constant intensity within the object. With each object in the image assigned to one of a number of different classes, an overall report can be given. The objects to be considered have no predefined shape or size and, although one may expect to see straight edges and angles in a crystal, this is not a prerequisite for diffraction. This means there is much overlap in the values of the variables expected for the different classes. However, each attribute gives some information about the object in question and, although no single attribute can be expected to correctly classify an image, it has been found that a combination of classifiers gives very good results.