|M.Sc Student||David-Horvitz Vered|
|Subject||Internal Quality of Fruit by Mealiness Assessment|
|Department||Department of Agricultural Engineering||Supervisors||Professor Emeritus Itzhak Shmulevich|
|Professor Raphael Linker|
This research investigates the possibility of detecting mealiness in apples via destructive and non-destructive methods.
With regards to the destructive methods, the best results were obtained using the tensile test of a fruit ring (98% correct classification as compared to professional sensory panel).
Two non-destructive methods were investigated, namely acoustic response and low mass impact. For each sensor, several parameters characterizing the time-varying signal were extracted. Classification was attempted using (1) linear combinations of the parameters obtained from each sensor separately, (2) linear combinations of the parameters obtained from both sensors, and (3) a neural network based only on the parameters extracted from the impact sensor. Analyzing each sensor separately led to correct classification of 90% and 80% for the impact and the acoustic device, respectively. Slightly better classification (94%) was obtained when the classification was based on both devices simultaneously. The neural network classifier led to correct classification of 90-93%.
In order to verify that the apples were indeed classified according to mealiness and not firmness, the analysis was repeated on a sub-sample containing only apples of similar firmness (similar E modulus). Again, good separation between mealy and juicy apples was achieved (91%-93% correct classification), indicating that mealiness was indeed detected.
The main achievements of this work are the establishment of the fruit ring tensile test as a reference method for mealiness determination, and the indication that, for mealiness assessment, the low mass impact test should be preferred to the acoustic test.