Image analysis of foliar greenness for quantifying relative plant health

Abstract

Foliar color has been used by farmers and specialists as visual health indicators for plants. However, assessment of health thru naked eye is subjective; hence, Munsell Plant Tissue Color Chart, Globe Plant Color Chart, and Leaf Color Chart have been developed. More accurate measurements of plant physiological processes such as leaf development, leaf senescence, light absorption, and foliar nutrient status, require the use of Chlorophyll meters, spectrophotometers, reflectometers and spectroradiometers. The aim of this work is to develop an experimental set-up to validate the use of digital images of foliar parts to quantify relative plant health of tomato plants. With this study, quantifiable measure can be achieved to rate the relative health of any plant by taking images of its leaves. To get reliable results, the model characterization used features extracted from captured images of sample leaves taken from 40 plants. Twenty plants are identified by a plant specialist as healthy and the other 20 plants as unhealthy. Bartlett’s and F-Tests were used to develop a statistical model which defines the group of healthy plants from the unhealthy ones. Randomly selected plants are then subjected to the model for classification and relative valuation of plant health.

Results show that the model successfully classified healthy plants against unhealthy plants using image analysis of foliar greenness. Twenty healthy and 20 unhealthy plants were selected by the plant specialist and then subjected to the model for classification. Thus, measures of dispersions between the pixels of foliar image as well as that of between leaves in three sections of a mother plant can be used to identify the relative health of plants.