A vision based crop monitoring system using segmentation techniques

Abstract

The characterization of health status for a plant using a non-destructive method is one of the challenging problems. In this study, the number of leaves and discoloration properties have been estimated using the images obtained from nine saplings of Solanum melongena (eggplant or brinjal) grown in the laboratory. The images were obtained using a mobile phone camera fitted on an automated device. A particle wave algorithm and contour grow technique was used for the segmentation of leaves which resulted in a segmentation accuracy of 89%. The defective percentage was estimated based on which saplings were ranked. Validation of healthy and defective regions was done by applying linear regression analysis on the estimated Normalized Green Red Difference Index (NGRDI) from images obtained using an automated device and a Foldscope (new paper-based microscope). The analysis resulted in R squared value and Least Mean Square Error (LMSE) of 0.86 and 0.1 respectively. © 2020 AECE.