Determination of brix value in tomato fruits using machine learning technique

Objective: The objective of this study is determination brix value in tomato by using machine learning technique. Materials and Methodology: Joker f1 tomato variety was used in the experiments of tomatoes. The color of the tomatoes was measured of tomatoes from the stem, middle and base regions in the CieL,a,b color space. The brix value was measured with a refractometer. 9 features were reached using 3 different measurements and 3 L,a,b color space parameters; namely L,a,b. Thus data set was consisted of color and brix values. Eta feature selection algorithm was used to determine significance of each feature on label. Support vector machines and decision trees are used as machine learning algorithms. Results: The brix value estimation model has a performance of R 0.437 (Decision Trees; 50% training 50% testing) for all colors. Conclusion: Model performance was poor in all models studied for brix estimation from color values.