Strawberry Volume Estimation Using Smartphone Image Processing

This study was performed to estimate the volume of strawberries using changes in the regularityof seeds on the fruit surface. The pwelch function in MATLAB was applied to images ofstrawberries and the proposed model was validated through correlation analysis between thecalculated volume and the actual volume of strawberry fruits. The volume estimation formulawas also used to derive a weight estimation formula. Therefore, an important part of theexperiment was to completely remove the background of images of strawberries so that only theseeds remained. The pwelch function was used to determine the periodicity of the strawberryseeds, but the remaining background affected the accuracy of results. As such, random noiseand image enhancement were applied to process the images for analysis. The periodicity of thestrawberry seeds was calculated for the x- and y-axes, and a positive correlation of y = 1067.9x+ 8.2903 (R² = 0.8946) was established between seed periodicity and fruit volume. However,there was a negative correlation of y = −4257.7x + 42.722 (R² = 0.8346) between seedperiodicity and the fruit volume. Because the two-dimensional images do not capture the threedimensionalityof actual strawberries, the distance between seeds tended to become smallertowards the ends along the y-axis. The calculated volume and the actual volume were comparedusing correlation formulas for the x-axis, y-axis and the average of the two axes. The coefficientof determination was 0.7439 for the x-axis, 0.8662 for the y-axis and 0.6027 for the average ofthe two axes. While both the x- and y-axes showed a high correlation between actual volumeand estimated volume, the formula in the y-axis showed a negative correlation. Thus, this studyconcluded that y = 1.7271x − 10.103 (R² = 0.8662) was the most suitable formula in estimatingthe volume of strawberries. KCI Citation Count: 0