Cotton hail disaster classification based on drone multispectral images at the flowering and boll stage

Abstract Traditional artificial statistical methods of hailstorm damage assessment are often inefficient, subjective, time-consuming, and inaccurate. The use of drones for high-resolution remote sensing imaging enables large-scale, rapid assessment of a target area. The purpose of this study was to use multispectral imaging techniques to promptly and accurately assess the damage degree of hail in crops. In this experiment, cotton was used as the research target, and the experiment was carried out after a hail disaster that occurred during the flowering and boll stage of cotton development. A hail vegetation index (HGVI) was constructed and compared with the classification results of other vegetation indices on the hail disaster degree. It was concluded that: (i) both the hail vegetation index HGVI and the previous vegetation index, ratio vegetation index, are suitable tools of classification for grading cotton hail disasters, with both indices having a Kappa coefficient over 0.85 and a median Kappa coefficient greater than 0.9; (ii) the method used in this study is suitable for simple classification based on multispectral images; (iii) this study explains the possibility of classifying hail damage degrees based on multispectral images. Subsequent research can further diagnose and classify hail disasters during different periods and improve the accuracy and completeness of classification.