UAV technology and machine learning techniques applied to the yield improvement in precision agriculture

Abstract

A model to estimate Nitrogen nutrition level in corn crops (Zea mays) is presented. The model was based on the information provided by multi-spectral cameras in four bands (red, green, blue and near-infrared (808 nm). The model was validated with ground truth information obtained by destructive methods. For training phase, three different fertilization levels of the crops were used (70, 140 y 210 kg · N · ha/sup -1/) with three repetitions in two stages of growing (V10 and earring). Unmanned Aerial Vehicle (UAV) technology was used. UAV quad-copter type flying 70 meters above the crops and machine learning techniques were used for the prediction stage. Results shown that the model can estimate nitrogen levels with 80% of precision with low cost technologies (multi-spectral cameras and UAVs). This proposal aims to optimize the fertilization since it actually is applied uniformly in the crops. The proposed scheme is focused on areas where the nitrogen is insufficient, avoiding the waste and reducing the impact on the environment.