Tomato Yield Estimation Based on Object Detection

Abstract

At present, the vegetable yield estimation in China is performed by manual sampling and visual observation of vegetable counts. This is not only time-consuming and labor-intensive, but it also has low precision. In this study, we capture video surveillance images of the tomatoes during plant maturation, and use neural networks to identify pictures, extract growing features, identify the number of vegetables hanging from the plants, and establish an estimation model for tomato yield. We then take a sample of the vegetables to be measured. Strains are image-analyzed and processed to predict yield per plant and yield per unit area to obtain an accurate prediction of tomato yield.