Application of UAV multispectral imaging for determining the characteristics of maize vegetation
¹University of Helsinki, Faculty of Agriculture and Forestry, Department of Agricultural Sciences, Koetilantie 5, FI00790 Helsinki, Finland
²Natural Resources Institute Finland (Luke), Latokartanonkaari 9, FI00790 Helsinki, Finland
³University of Helsinki, Helsinki Institute of Sustainability Sciences (HELSUS), Yliopistonkatu 4, FI00100 Helsinki, Finland
⁴University of Helsinki, Faculty of Agriculture and Forestry, Department of Agricultural Sciences, Latokartanonkaari 5, FI00790 Helsinki, Finland
*Correspondence: mikael.anakkala@helsinki.fi
Abstract:
Interest in forage maize (Zea mays L.) cultivation for livestock feed has grown in northern conditions. In addition, it is important to develop methods and tools to monitor crop development and other characteristics of the crop. For these purposes UAVs are very efficient and versatile tools. UAVs can be equipped with a variety of sensors like lidar or different types of cameras. Several studies have been conducted where data collected by UAVs are used to estimate different crop properties like yield and biomass. In this research, a forage maize field experiment was studied to examine how well the aerial multispectral data correlated with the different properties of the vegetation. The field test site is located in Helsinki, Finland. A multispectral camera (MicaSense Rededge 3) was used to take images from five spectral bands (Red, Green, Blue, Rededge and NIR). All the images were processed with Pix4D software to generate orthomosaic images. Several vegetation indices were calculated from the five spectral bands. During the growing season, crop height, chlorophyll content, leaf area index (LAI), fresh and dry matter biomass were measured from the vegetation. From the five spectral bands, Rededge had the highest correlation with fresh biomass (R2 = 0.273). The highest correlation for a vegetation index was found between NDRE and chlorophyll content (R2 = 0.809). A multiple linear regression (MLR) model using selected spectral bands and vegetation indices as inputs showed high correlations with the field measurements.
Key words:
agriculture., maize, multispectral images, UAV, vegetation index