Multiple Linear Regression models for estimating the plant height of red pepper using UAV-based multispectral imagery

Abstract

Red pepper (Capsicum annuum L.) is a major field crop in Korea; however, its productivity and cultivation area have recently declined due to labor shortages caused by an aging population and climate variability. This study was conducted on the ‘Colormura’ cultivar, transplanted on April 29 in a farm field located in Bangyo-ri, Gimje-si. The objective was to develop a Multiple Linear Regression (MLR) model to estimate plant height—a key growth indicator highly correlated with yield—using vegetation indices derived from Unmanned Aerial Vehicle (UAV)-based multispectral imagery. Plant height data and UAV multispectral images were collected at three time points: mid-June, mid-July, and mid-August. To evaluate the model's generalization performance, the data were partitioned into calibration and validation datasets at ratios of 8:2, 7:3, and 6:4, and performance was assessed using R2, RMSE, and MAPE. Because individual monthly analyses resulted in low generalization performance or overfitting, the model was constructed using time-series data spanning from June to August. The model utilizing PRI, TCARI, NDRE, and OSAVI exhibited the best performance. Notably, the RedEdge-based indices, NDRE and TCARI, made significant contributions to the prediction of plant height. This study demonstrates the potential of estimating red pepper growth using UAV multispectral data. Future research should focus on enhancing model performance by expanding the dataset and applying diverse analytical techniques.