Maturity and Yield Estimation of Tomatoes Using RGB and Multispectral Images

Due to the rising cost of labor, smart agriculture is regarded as the trend of future agricultural development. In this work, we propose a system for maturity and yield estimation of tomatoes using RGB and multispectral images. The tomatoes are identified using object detection and tracking networks, followed by ripeness classification into the categories "mature", "almost mature" and "immature". In the experiment, the image sequence is used to detect tomatoes, and assign an ID to each of them. The principal component analysis is then performed using multispectral indices including NDVI, GRRI and GNDVI. Finally, tomato tracking and maturity classification are integrated, and the number of tomatoes with various maturity in each group is counted. Through this research, we can more effectively monitor and manage tomato growth and improve agricultural production efficiency.