A UAV-based framework for crop lodging assessment

Abstract

Crop lodging assessment needs to be carried out timely and accurately to ensure valuable information about the location and area where lodging occurs. Many applications have been explored and tested for unmanned aerial vehicle (UAV) visible imagery in agricultural management due to the ability of providing high-space-resolution information. However, there still face many challenges in extracting lodging information using UAV visible imagery, and lacks consensus on an appropriate way to assess crop lodging. The main purpose of this study was to proposed an efficient framework to identify crop lodging at the field scale using UAV visible imagery. This framework contained a two-phase procedure. Meanwhile, three methods were evaluated by providing the appropriate feature subset for objected-based classification to determine the best feature selection method. The results showed that the proposed framework provided high accuracy (94.0 %) for identification of sugarcane lodging. Furthermore, the Boruta algorithm yielded the best feature subset compared with statistical indicators and the RFE algorithm. Thus, the proposed framework based on UAV visible imagery is promising to identify crop lodging precisely, and has great application potential in precision agriculture.