Comparison and Evaluation of Vegetation Indices for Image Sensing Systems in Precision Agriculture

In precision agriculture, remote estimation processes such as the determination of the canopy area, the yield and plant volume are useful for the camera based observations. Especially, using accurate vegetation index for identifying the plants is very important in the remote image sensing systems. In this study, four frequently used Red Green Blue (RGB) vegetation indices were compared to the manually extracted plant region images of interest obtained in the ImageJ software by using Otsu thresholding method. Excess Green (ExG), Excess Green minus Excess Red (ExG-ExR), Green Percentage (G%) and Triangular Greenness Index (TGI) indices were used with digital color images of single artificial plants. A novel image processing algorithm was developed in LabVIEW software for determining the green area of artificial plants. Performances of mentioned RGB vegetation indices were also evaluated and compared with each other. The comparison of mentioned vegetation indices showed that the highest congruence ranking had been statistically obtained for the ExG and TGI indices at the estimations of the canopy areas according to the vegetative index. While both the ExG and TGI indices presented more proximities to the reference point (mean pixel values obtained by using ImageJ software), ExG-ExR and G% indices showed worse proximity performances to the referredmean pixel values obtained by using ImageJ software. According to the results, both ExG and TGI indices showed reliable separability and can be preferred for green canopy area estimation in image sensing systems.