Abstract
Crop height especially for rice crops is an important agronomic
parameter that enables calculations of biomass, yield, and plant
nitrogen use, as well as the assessment of lodged areas. Traditionally,
crop height has been measured with a ruler, however, this method is
time-consuming, ineffective, and subject to human error. Using
vegetation indices (VIs) produced from remote sensing data is an
indirect approach to the estimation of crop height. VIs is not sensitive
to variations in crop height during the later growth stage. Recent
research has used a variety of remote sensing platforms to measure crop
height. In this study, a multispectral (MS) imagery-based digital
surface model (DSM) was used for crop height estimation. A field
experiment was carried out in the Agricultural and Food Engineering
(AgFE) Department, IIT Kharagpur. For this experiment rice crop
(Variety: Super Shankar) was transplanted in 6 plots of 10m×10m size and
4 plots of 10m×10m size were kept barren. Crop height was measured at
every 10 days interval and MS imageries of the same date were acquired
using a quadcopter UAV. Pix4Dmapper Pro software was used for
ortho-mosaicing the MS imageries and the generation of the DSM.
Generated DSM was processed in ArcGIS and Extracting DSM for crop plots
as well as barren plots. Ground pixels were separated from the canopy
pixels by assuming the threshold values. Crop height is calculated by
subtracting pixel values of the canopy from the pixel value of bare
land. DSM-based crop height estimation yielded an RMSE of 6.3 cm for the
crop growth period. This approach to crop height estimation may be
helpful for the calculation of important agronomic and phenotypic
parameters.
Keywords: Crop Height, UAV, Multispectral Imageries, Digital
Surface Model (DSM)