Estimating leaf area index of sugarcane based on multi-temporal digital images

Sugarcane is one of the most important economic crops in south China. It's of practical value to quickly assess the leaf area index (LAI) of sugarcane and evaluate its growing state. Field-based or in situ measurements are very labor intensive and time consuming, while satellite images are relatively coarse and unsuitable for real-time crop monitoring, not to mention the cloud-prone climate in south China. In this study, an automatic image processing technique is proposed for the estimation of sugarcane LAI using digital images. The test site is a sugarcane plantation in Guangxi province, southern China. In our experiment, a digital video camera was fixed on a 6.0 m pole to acquire sugarcane images during the entire sugarcane growing season. Reference data were also collected through field-based LAI measurement using a portable handheld leaf area meter. The daily acquired sugarcane images were processed in a MATLAB environment and the different color vegetation indices were extracted from the time series digital images. The linear relationship between different color vegetation indices and field-measured sugarcane LAI values were established. The color vegetation indices with the highest correlation coefficients with field-measured LAI were chosen to fit and test the model against the ground-measured LAI values. Based on the root mean square error (RMSE) and coefficient of determination R
²
between field-measured LAI and model-predicted LAI, we concluded that the most suitable color index for sugarcane LAI estimation during the complete growth period was G-B, and the most appropriate image shooting time for use in the construction of the estimation model for sugarcane LAI was confirmed and also proposed for further studies.