Evaluation of different model approaches for mass estimation of apple fruit using non destructive impact technique

In this study, it is aimed to estimate the mass of apples by using the nondestructive impact technique and to develop different model approaches. Starkrimson apple varieties were used in the experiments. In the estimation of apple mass, 10 nondestructive impact parameters were removed using force-time curves, and stepwise (step-by-step) regression analysis method was used to reduce the number of impact parameters to be used in the mathematical model (Fmax1, tmax, tmax1, Ia and tP1-2). Apple mass estimation was made by using these parameters in the multiple linear regression analysis method (MLR). According to the results of statistical analysis, developed mathematical model estimated the apple mass with 3.07 g and 3.35 g prediction error in the calibration and verification data groups, respectively. In the calibration and verification datasets, determination coefficients of the apple mass estimation (R2) were calculated as 0.94 and 0.93, respectively. The success of the mass estimation model according to the mass groups classified according to the cluster analysis was also determined. According to the results of the data group analysis, the true accuracy of the model approach was calculated as 32. In addition, the success of the classification of apple samples was calculated as 94.11%.