An Automatic Cognitive Vision Approach For Determination Of Growth And Quality Of Betel Leaves

Abstract

Growth and quality analysis of any plant provides physiological and quantitative interpretation of the performance of any plant. In commercial crops, such as Betel leaf has cultural and economic significance where farmers earn their livelihood through its cultivation, quantitative plant growth and quality assessment is necessary for yield management. Existing growth and quality analysis systems make use of manual approach that are destructive, expensive, time consuming and labor intensive. Remote sensing and Ariel imaging is not possible in case of betel leaf cultivation and so very few researchers have reported its growth and quality analysis. Therefore betel leaf is considered as plant material in the proposed research work. Leaf samples were collected from three different states, namely Chhattisgarh, West Bengal, and Orissa. newlineNo research has been reported till date for determining growth and quality of betel leaf. The proposed research uses machine vision approach to estimate physiological parameters such as leaf area, dimension, chlorophyll content, and disease status to quantify the growth and quality assessment of betel leaf. In the proposed work leaf area has been estimated through image processing approach and also correlated with traditional grid counting, and paper weight method. A mathematical model has been developed to predict SPAD and Chlorophyll concentration from corresponding RGB values. Identification of leaf disease has been done through its color, geometrical and morphological features. A NN model has been developed for betel leaf disease classification. A significant linear correlation (R2 = 0.9576) has been observed between traditional and proposed approach for leaf area measurement. Also, significant correlations (R2 = 0.9562) has been observed between chlorophyll meter (SPAD) readings and proposed RGB color model. The proposed NN model used for disease classification has resulted in an average accuracy of about 95.3%. newlineTherefore, the proposed machine vision technique results in an automatic,