Application of Artificial Neural Networks to Estimate Color Surface Features of Three Maturity Stages of Tomato Based on Dimensions and Weight

Abstract

Aims: The aim of this research was to investigate the effect on application of artificial neutral network (ANN) to estimate the color surface of fruit of three maturity stages of tomatoes based on fruit dimensions (length and width) and weight. Study Design: Simple machine vision system was built to extract color surface features of tomato samples. Place of Study: Agricultural and Bio-systems Engineering Department, Faculty of Agriculture, Alexandria University, Egypt. Methodology: Samples of variety of tomatoes (Baladi variety) were manually harvested from the field at Educo, El- Beheira Governorate, Egypt. Three maturity stages of the variety were harvested in different days by eye inspection based on their color. The maturity stages were green, pink and red. The weight and dimensions of each maturity samples were measured. Samples images were taken on a white background and manual mode, no zoom, no flash were used by the camera. Surface color of the tomato samples was analyzed quantitatively. ANN model to estimate the surface color was applied. Results: The evaluation results of testing data set showed that ANN could be able to estimate color surface features of tomatoes at different accuracy as evaluated by coefficient of determination (R2) of 0.7161, 0.8273, 0.8605, 0.5448, 0.8056, 0.7954 and 0.854, respectively for L*, a* b* Hue, Chroma, color index and color difference with true red. The obtained weights from the ANN training process were formulated in Excel spreadsheet. Conclusion: The studied color surface features of tomato for three maturity stages and input variables well correlated. The tomato weight contributed significantly in estimating all surface color features of tomato compared to the length and width. The developed Excel spreadsheet could be used as a quick tool to estimate color surface features of tomato.