As one of the specialty crops of the Ningxia Hui Autonomous Region in China, Xisha watermelon is primarily cultivated in large open fields. Estimating yield distribution is essential for enhancing agronomic operations, evaluating breeding processes, and planning harvests and market distribution. However, yield estimation for Xisha watermelon is a seasonal and labor-intensive task. This paper proposes a global scanning yield estimation method based on drone remote sensing technology, which efficiently calculates the yield distribution of Xisha watermelon fields. Firstly, an individual watermelon weight estimation model was developed using an elliptical volume-to-weight regression, achieving a goodness of fit of 0.942, which demonstrates the feasibility of image-based yield estimation. Next, a training set for the watermelons was prepared, and the optimal detection weights were obtained using YOLOv8n, achieving an F1 score of 0.968. Finally, field experiments were conducted in the target watermelon field to obtain 2D orthophotos. YOLOv8n and traditional machine learning methods were employed for target detection and elliptical fitting, and an image cropping and data processing technique was used to generate the yield distribution map. The global scanning yield estimation method described in this paper can be effectively applied to estimate yield distribution in Xisha watermelon fields.