Discrimination Analysis of Corn Species Using Field Hyperspectral Data

Abstract

One of eminent agriculture commodities in the world is corn. In Indonesia, corn is also important commodity both as staple food and export trade. Therefore, the practice of proper farming management is demanding as good employment of such practice would ensure the sustainability of corn production level. Understanding corn characteristics is one of important aspects for farming management, and to solve that task, hyperspectral technology is utilized. Hyperspectral technology is widely known for its ability to compensate the limitation of multispectral technology. Its numerous spectral resolutions make it possible to perform deeper and more comprehensive analysis than that of using multispectral. Spectrometer is a tool that uses hyperspectral technique to acquire spectral information. Using this tool, the characteristics of spectral reflectance from predominant corn varieties in Indonesia, Bisma and Pioner were successfully collected and their separability levels were measured using ANOVA One Way Tukey HSD statistic, Jeffries Matuista and Transform Divergence methods. These methods were able to conclude that region in Blue and Red visible were not suitable for discrimination as they showed dominant values in poor separability class, in the other way, region green and NIR were the best region with dominant values in good and perfect separability classes. From different perspective, which is from their phenology stages based on Vegetation Indice NDVI, the best performing stage for discrimination was flowering stage following its highest Jeffries – Matuista and Transform Divergence values and the lowest p-value among others. Meanwhile the worst stages were vegetative and yield ripening formation