Color and Grey-Level Co-Occurrence Matrix Analysis for Predicting Sensory and Biochemical Traits in Sweet Potato and...

Area of Science:

• Agricultural Science
• Food Science
• Biotechnology

Background:

• Sensory traits are crucial for consumer acceptance and market success of sweet potato and potato cultivars.
• Current sensory evaluation methods are time-consuming, potentially delaying breeding program progress.

Purpose of the Study:

• To evaluate the relationship between sensory panel assessments and instrumental color/texture features in sweet potato and potato.
• To explore the potential of instrumental analysis as a faster alternative for sensory trait evaluation in crop breeding.

Main Methods:

• Correlation analysis between sensory panel data and instrumental color/texture measurements.
• Development and evaluation of machine learning models (XGboost, NL-SVM) using instrumental data to predict sensory traits.
• Screening of chemical properties (starch) in sweet potato for correlation with sensory and instrumental features.

Main Results:

• High correlations (up to r=0.84) were found between sensory panel color scores and instrumental color measurements in both crops.
• Moderate correlations (up to r=0.66) were observed between sensory aroma/flavor and instrumental color, requiring further validation.
• Machine learning models showed moderate accuracy in predicting sensory traits, with XGboost and NL-SVM performing comparably (r^2=0.64-0.72 for aroma/flavor).
• Starch content in sweet potato moderately correlated with mealiness (r=0.54) and instrumental color (r=0.65).

Conclusions:

• Instrumental color analysis is a viable and efficient alternative to sensory panels for color scoring in sweet potato and potato.
• Instrumental methods, coupled with machine learning, can aid in the initial selection of genotypes for aroma, flavor, and texture.
• Further research is needed to elucidate the complex relationship between instrumental color and sensory attributes like aroma and flavor.