FUZZY LOGIC MODEL FOR PREDICTING PEANUT MATURITY

Abstract

Peanut quality and yield are impacted by harvest timing. The most commonly used tool for determining harvest timing is the hull-scrape chart giving association of kernel maturity to color of mesocarp. The hull-scrape technique is tedious, time-consuming, and labor-intensive. Wider use of maturity evaluations would be greatly facilitated by a quicker and easier test. While testing for maturity, the NMR signals from peanuts and days after planting exhibit a nonlinear relationship with the maturity class of kernels. Therefore, linear classification techniques such as linear discriminant analysis (LDA) may not achieve “good” classification results. This article describes the development of a fuzzy model to predict peanut maturity based on NMR-signal (FIDPK) and days after planting (DAP). Compared to the hull-scrape method, the fuzzy model predictions were 45%, 63%, and 73% accurate when maturity was classified in 6 classes, 5 classes and 3 classes, respectively. The respective accuracies from LDA, using the same data, were 42%, 56% and 70%. Data from 346 kernels were used for performance evaluation of both the fuzzy and LDA models. The fuzzy model improved maturity prediction compared to LDA. These results are encouraging, however, fuzzy model should be further evaluated with new data.