SPATIAL AND SPECTRAL CHARACTERISTICS OF CORN LEAVES COLLECTED USING COMPUTER VISION

Abstract

Current crop modeling techniques are unable to thoroughly describe plant architecture. The primary parameters of interest in corn canopy growth and development are leaf area and physiological state. Past researchers have developed equations relating leaf length, width, and area. This study attempts to update such relationships using data collected on 963 leaves from 87 separate com plants sampled from the field at given times during the season. Spatial and spectral parameters were recorded from slide images and were analyzed using a high resolution color vision system. Leaf areas were found to be highly correlated with leaf lengths. Additionally, when length and width were combined, area predictions were improved. Plant front views were investigated to determine the feasibility of estimating leaf areas from two dimensional leaf lengths. Leaf lengths measured were highly correlated with actual vision system leaf lengths (r^ = 0.99). Individual leaf areas predicted using measured lengths were highly correlated (r^ = 0.95) with actual areas. When combining areas to determine a total leaf area value for a given plant, the r^ value increased to 1.00. The physiological state of a leaf could also be assessed using its average color. Senesced leaves could be statistically separated from living leaves with a minimum confidence level of 95% using red and green color planes.