Crop type discrimination through low cost proximal RGB imaging and multivariate analysis

Abstract

Abstract The current study is an attempt to use low cost red green blue (RGB) image based vegetation indices (VIs), obtained from simple RGB camera, in separating six different field crops. To achieve this, sixteen common VIs were calculated and used as inputs in different multivariate analysis for separating wheat (Triticum spp), mustard (Brassica spp), cabbage (Brassica oleracea), pigeon pea (Cajanus cajan), brinjal (Solanum app) and chickpea (Cicer arietinum). Based on the variation in the green red ratio index (GRRI), Colour intensity index (INT), Color index of vegetation (CIVE) and Woebbecke index (WI) were identified performing significantly (p < 0.05) in discriminating six different crops e.g., cabbage, wheat, mustard, brinjal, pigeon pea, chick pea through classification and regression tree (CART) analysis. The results obtained from CART analysis were subsequently compared with discriminant analysis, which showed an accuracy of 96.3% of classifying different crops. The study meaningfully identified sensitive VIs that can be used to classify different field crop. The information achieved in this study can help in commercial and scientific decision making, planning in agribusinesses, and can be an important tool for conducting crop survey at regional scale.