Firmness prediction and modeling by optimizing acoustic device for watermelons

Firmness can be used to indicate ripeness of many agro-products, usually determined by acoustic impulse method nondestructively. An acoustic device was developed after investigating the influence of hitting ball and fruit tray on spectrum. Three firmness indices such as f2m, MI1 (index of the first order moment) and MI2 (index of the second order moment) were proposed to correlate with firmness of watermelon. Significant correlation was found out between firmness and these indices using linear regressive model and nonlinear model of artificial neutral network (ANN). It was concluded the linear model was more suitable than nonlinear model using ANN because of little difference of correlation coefficients. Although more computations needed, index MI1 was considered more precisely to use as firmness indices than f2m in the case of splitting peaks with almost the same amplitude around the first resonant frequency.