Modeling the Relationship Between Root Color, Root Shape, and Weight of Radish Using Random Forests

Abstract

The quality of radishes depends on the production environment and postharvest management throughout the supply chain. Quality monitoring is therefore important for post-harvest management in the supply chain. Here we estimate the quality of radishes by non-destructive methods based on color and shape using a random forest algorithm in a data-driven predictive model. The explanatory variables (i.e., color and shape) were obtained by capturing images of radish under a controlled photographic environment. Color information was converted from RGB to HSL or HSV to minimize potential effects of light conditions on the radish surface. Model performance was assessed using Pearson’s correlation coefficient (COR), Nash-Sutcliffe efficiency (NSE), and root mean squared error (RMSE). Experimental results indicate high model performance, supporting the applicability of non-destructive weight estimation using only radish color and shape. Among the models using different color components, the HSV model exhibited the best result of all performance measures in this study (i.e., COR, NSE, and RMSE were 0.889, 0.776, and 1.55, respectively). In accordance with the radish variety in this study, the R value was the most important variable among all color components. Partial dependence plots provided more detailed visual relationships between each unique pair of color components and radish weights. This method is promising for application in other photographic environments, specifically where sunlight is present. Further studies on the assessment of internal conditions of fresh radish such as vitamin C and anthocyanin could be useful for consumer-oriented quality assessment in future.