Premium
An Objective Dry Pea ‘Colour’ Scoring System for Commercial and Plant Breeding Applications
Author(s) -
Coles Graeme D
Publication year - 1997
Publication title -
journal of the science of food and agriculture
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.782
H-Index - 142
eISSN - 1097-0010
pISSN - 0022-5142
DOI - 10.1002/(sici)1097-0010(199708)74:4<435::aid-jsfa807>3.0.co;2-0
Subject(s) - ranking (information retrieval) , chrominance , statistics , mathematics , calibration , set (abstract data type) , sample (material) , rank (graph theory) , standard deviation , scoring system , training set , sample size determination , artificial intelligence , computer science , luminance , medicine , surgery , chemistry , chromatography , combinatorics , programming language
For human consumption, colour of dry peas is an important quality criterion, presently assessed subjectively. Samples which are dark and even in colour are preferred to those which are bleached and variable. The capacity of image analysis methods to provide an objective colour scoring system was investigated, and useful predictive relationships were found between subjective rankings for colour allotted to members of a large set of training samples, and ranking for median optical density, optical density proportional range and median seed size. Training samples were drawn at random from the pea sample population, and were not evenly distributed, so spacings between rankings were corrected by reference to the variance of the mean ranking for a sample. This correction allowed the predictive variables to explain 64% of the observed variation in the corrected rank. When the required precision of a derived scoring system was set to ±1·0 score points, it was found that the set of training samples could be covered by 6·95 score points, so a scoring calibration equation was developed in which increasing colour loss caused a higher score. The darkest sample in the training set was arbitrarily chosen to have a score of 5, to allow for the possibility of intrinsically darker, less bleached samples. This scoring system is at least twice as precise as the current subjective approach, and can be applied using any image analysis system that can be calibrated against cheap, highly consistent, photographic standard grey cards. The performance of the system may be improved further if chrominance variation can be taken into account. © 1997 SCI.