Click on “Download PDF” for the PDF version or on the title for the HTML version.

If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.

Abstract. Soybean varieties are generally sensitive to flooding stress. Under flooding stress, soybean plants with different flooding tolerant abilities show different levels of morphological and physiological responses. Traditionally, these levels of responses are measured by experienced researchers and grouped into several categories, e.g. flooding injury rating (FIR) from 1 to 5 as 5 is the severest injury and 1 is the slightest injury. The goal of this study is to develop a flooding injury classification method using UAV-based imagery and machine learning methods for in-field soybean flooding experiments. A total of 396 soybean genotypes were exposed in flooding stress in 2019. UAV-based multispectral and infrared thermal images were taken Sept. 6, 2019. The extracted features such as NDVI, RVI, SAVI, GLI, NDRE and CA are optimized by Pearson Correlation Coefficient (PCC) to compose a feature database and visual observed FIR were used as ground truth data. Finally, the support vector machine (SVM) classifier was used for classifying. This proposed classifier system is capable to classify the types of FIR of soybean with a high accuracy, indicating that the proposed method has great potential in qualifying soybean responses to flooding stress.