Monitoring of maize chlorophyll content based on multispectral vegetation indices

Abstract

In order to estimate the nutrient status of maize, the multi-spectral image was used to monitor the chlorophyll content in the field. The experiments were conducted under three different fertilizer treatments (High, Normal and Low). A multispectral CCD camera was used to collect ground-based images of maize canopy in green (G, 520~600nm), red (R, 630~690nm) and near-infrared (NIR, 760~900nm) band. Leaves of maize were randomly sampled to detect the chlorophyll content by UV-Vis spectrophotometer. The images were processed following image preprocessing, canopy segmentation and parameter calculation: Firstly, the median filtering was used to improve the visual contrast of image. Secondly, the leaves of maize canopy were segmented in NIR image. Thirdly, the average gray value (G_IA, R_IA and NIR_IA) and the vegetation indices (DVI, RVI, NDVI, et al.) widely used in remote sensing were calculated. A new vegetation index, combination of normalized difference vegetation index (CNDVI), was developed. After the correlation analysis between image parameter and chlorophyll content, six parameters (G_IA, R_IA, NIR_IA, GRVI, GNDVI and CNDVI) were selected to estimate chlorophyll content at shooting and trumpet stages respectively. The results of MLR predicting models showed that the R² was 0.88 and the adjust R² was 0.64 at shooting stage; the R² was 0.77 and the adjust R² was 0.31 at trumpet stage. It was indicated that vegetation indices derived from multispectral image could be used to monitor the chlorophyll content. It provided a feasible method for the chlorophyll content detection.