Estimation of Peanut Seedling Emergence Rate of Peanut Based on Deep Learning and Drone Imagery

Obtaining the number of peanut seedlings and seedling emergence rate in a timely and accurate manner is of great significance for yield prediction. However, current labor survey methods of peanut seedling have problems with backward means, time-consuming and inefficient. In this study, peanut seedlings collected from various peanut germplasm resources and subjected to different management conditions during 2022–2023 were used as the research subjects. Visible light images of the peanut fields, captured by unmanned aerial platforms, served as the data source. A deep learning method was employed to construct a training model for detecting peanut seedlings. The results show that the model trained based on the YOLOv5-1 algorithm performs best in peanut seedling number recognition, with a precision of 98.34% in the validation set, a recall rate of 96.09%, and an average precision (AP) mean of 97.98%. The model shows good applicability in complex field environments and can quickly and accurately detect the number of peanut seedlings during the peanut seedling stage, which provides a technical reference for peanut seedling counting and yield estimation in the field.