Development of interactive multiple models for individual fruit mass estimation of tomatoes with diverse shapes

• Establishing interactive estimation models for fruit weight of tomato with ten different shapes. • Developing an online system for non-destructive fruit weight estimation of on-vine tomatoes. • Improving the intelligence level of dynamic monitoring for the fruit weight of on-vine tomatoes. Dynamic monitoring of tomato fruit size is critical for precision cultivation management. This study aimed to develop accurate and efficient models for the non-destructive in situ detection of individual fruit mass (IFM) across 10 common tomato shapes. Using 2044 ripe fruits, we established mathematical models to estimate the IFM based on transverse diameter, longitudinal diameter, and 14 diameter-derived parameters (DDPs). The modelling process involved data analysis approaches and an interactive sequential calculation series for various tomato shapes. The optimal shape-specific IFM models achieved determination coefficients of 0.965–0.988, with mean absolute percentage errors (MAPE) < 9.77 % and mean absolute errors (MAE) < 7.50 g for all non-flattened shapes. Applicability tests demonstrated that the MAPE and MAE of those optimal models were below 20 % and 4.34 g respectively for non-flattened fruits, confirming their high accuracy and field applicability. The mini-program and interactive modeling system were further developed based on those fruit-shape-specific models, significantly enhancing the practical application of the study’s findings and offering a rapid and precise estimation method for in situ detection and dynamic monitoring of IFM in diversely shaped tomatoes, as well as for yield prediction.