Development of a Digital Analysis System to Evaluate Peanut Maturity

Abstract

The profile color class method developed by Williams and Drexler in 1981 for the prediction of peanut harvest has proven to be a relative description of peanut maturity and is currently used by growers. However, the method requires the subjective visual classification of pods based on the development of color in the mesocarp layer of the hull which naturally introduces variability and possible error in maturity prediction based solely on observer bias. A Digital Image Model (DIM) was developed to eliminate subjectivity in pod color classification. The DIM is a method in which a scanned image of pod mesocarp colors is analyzed using a color definition algorithm. The final output of the DIM is a ratio of pixel color classes. To develop the DIM, replicated plots were established in Florida in 2010 and 2011 and sequentially harvested starting at 120 days after planting (DAP) and then progressing at wk intervals through 155 DAP. At harvest, yield and grade were evaluated for each plot and pod samples were collected for color classification by a single observer using the current profile board method. These same pod samples were then imaged and analyzed with the DIM method. The percentage of black and brown pods (mature pods) classified by the profile board and the DIM method were evaluated to determine the overall performance of the DIM in comparison to the profile board. The DIM method was successful in predicting the percentage of black and brown pods similarly to the profile board in both years with R2 0.63 to 0.82 with images acquired from the saddle region of the pod. There was more variability in matching the DIM prediction to the profile board when imaging pods from random regions, with R2 0.19 to 0.82. The goal of this research was to develop an imaging system that could be accessed by growers, consultants, and extension agents for objective analysis and prediction of peanut maturity.