A novel tobacco-oriented vegetation index from multi-spectral remote sensing imagery

Remote sensing vegetation index (VI) is a simple, effective, and widely used approach for quantitative and qualitative assessment of vegetation vigour, coverage, activity, and growth dynamics. To address the insufficient separability and high background sensitivity of existing VIs in tobacco identification, this study proposes a tobacco-oriented vegetation index (TVI). The TVI formulation incorporates four factors: (1) the high reflectance characteristics of tobacco leaves in the green band; (2) the pronounced difference between green band and red band among various land-cover types; (3) the pronounced difference between green band and blue band among various land-cover types; (4) and the canopy enhancement effect induced by the accumulation of visible-light brightness. Experimental results demonstrate that TVI exhibits high consistency and robustness in automatic flue-cured tobacco identification and area extraction, achieving an average extraction accuracy of 85.01% and an average extraction rate of 94.76%, and mean Precision, Recall, and F1 values of 85.01%, 89.80%, and 87.26%. Overall, TVI, with its clear physical rationale, concise formulation, and stable performance, provides a practical and reliable remote sensing tool for tobacco identification, area estimation, and growth monitoring.