Discriminant Analysis of Defective and Non-Defective Field Pea (Pisum sativum L.) into Broad Market Grades Based on Digital Image Features

May Discriminant Analysis of Defective and Non- Defective Field Pea (Pisum sativum L.) into Broad Market Grades Based on Digital Image Features Linda S. McDonald 0 1 Joseph F. Panozzo 0 1 Phillip A. Salisbury 0 1 Rebecca Ford 1 0 Agriculture Research, Department of Economic Development , Jobs, Transport and Resources, Horsham, Victoria , Australia , 2 Faculty of Veterinary and Agricultural Sciences, The University of Melbourne , Parkville, Victoria , Australia , 3 School of Natural Sciences, Griffith University , Nathan, Queensland , Australia 1 Editor: George-John Nychas, Agricultural University of Athens , GREECE Field peas (Pisum sativum L.) are generally traded based on seed appearance, which subjectively defines broad market-grades. In this study, we developed an objective Linear Discriminant Analysis (LDA) model to classify market grades of field peas based on seed colour, shape and size traits extracted from digital images. Seeds were imaged in a highthroughput system consisting of a camera and laser positioned over a conveyor belt. Six colour intensity digital images were captured (under 405, 470, 530, 590, 660 and 850nm light) for each seed, and surface height was measured at each pixel by laser. Colour, shape and size traits were compiled across all seed in each sample to determine the median trait values. Defective and non-defective seed samples were used to calibrate and validate the model. Colour components were sufficient to correctly classify all non-defective seed samples into correct market grades. Defective samples required a combination of colour, shape and size traits to achieve 87% and 77% accuracy in market grade classification of calibration and validation sample-sets respectively. Following these results, we used the same colour, shape and size traits to develop an LDA model which correctly classified over 97% of all validation samples as defective or non-defective. - OPEN ACCESS Funding: This project was supported by the Grains Research and Development Corporation through grant DAV00132 and the Department of Economic Development, Jobs, Transport and Resources. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Introduction Field pea (Pisum sativum L.) is generally traded based on broad quality grades, each of which has its own market niche. Grades are determined subjectively and often classified inconsistently between the grain sellers and buyers, leading to trading disputes. Khan and Croser [ 1 ] identified five broad types of field pea (yellow, marrowfat, dun, green/blue and maple) and six quality traits which heavily influence their marketing; admixture levels, insect damage, seed colour, seed size, seed cleanliness and product uniformity. Historically, these traits are based Competing Interests: The authors have declared that no competing interests exist. on appearance and are assessed visually. As such, the trading value of field pea (like most pulse grains) is subjectively determined. There is an opportunity, therefore, for objective measurement of grain products using colour grading systems or machine vision to reduce the potential for inconsistent assessment. Within the grains research field, several studies have been conducted on the application of machine vision systems to quantitatively determine characteristics related to grain quality. Zapotoczny and Majewska [ 2 ] investigated the measurement of wheat colour, of both the endosperm and grain coat, by machine vision. Fundamental size traits, such as grain length, width and volume, have been modelled in various studies [ 3–6 ], as well as shape of grains [ 7– 9 ]. Further to grain size, shape and colour analysis, machine vision studies have also been applied to assess traits which impact on grain processing, such as chalkiness in rice [ 10, 11 ], performance of wheat samples through a dockage tester (Paliwal, Visen et al. 2003) and distribution of grain size [ 12, 13 ], which impacts on milling efficiency. Machine vision sytems have also been used in the grains industry for colour-based grading and identifying defects and seed damage. While 2-dimensional colour, size and shape traits are the most commonly used, more recent focus has been on expanding the range of image traits to include textural, morphological, and wavelet features, enabling a suite of measurments from each image and contributing to an increased efficency and justification of the capital expenditure in setting up digital image technology. Anami and Savakar [ 14 ] provided a summary on some of the most common feature extraction methods used in the analysis of grains, fruits and flowers. Choudhary, Paliwal [ 15 ] developed a model to classify cereal grains into grain type (wheat, rye, barley and oats) and reported that the combination of morphological, colour, textural as well as wavelet features gave the best results for classification. A number of studies (...truncated)