Leaf area assessment using image processing and support vector regression in rice

Abstract

Crop growth, health, and correspondingly yield are much affected by abiotic environmental factors. Abiotic stress is considered as a threat to food security and has a disastrous consequence. Phenotyping parameters such as leaf area assessment is of utmost importance in determining the stresses due to water and environmental factors, micronutrients deficiencies, leaf diseases, pests, etc. In this study, a non-destructive approach through digital image analysis has been presented to assess the total leaf area of rice plants grown in pot culture. Images have been captured from four different angles with respect to the initial position of the camera. Support Vector Regression (SVR) and Tuned SVR have been employed by considering the pixel area of leaves obtained from different angles. Performance of Tuned SVR has been found better than the SVR on training and testing dataset based on RMSE values. A web-solution has been designed and developed to implement the presented approach using 3-tier architecture: Client-Side Interface Layer (CSIL), Database Layer (DL) and Server Side Application Layer (SSAL).