This paper studies mainly apples non-destructive selection technology, and presents the principle and implementation of non-destructive selection system for apple quality based on the theory of the fuzzy and neural network. Firstly, we can differentiate the kinds of apples by neural network theory. Then, the apple’s quality greatly depends on its size, shape, and color, which are considered as three input parameters in selection system. The pick-up of three character parameters is handled by computer image processing. Only one output parameter is apple rating, which is fall into three grades: A, B and C. Regarding three character parameters as input, we obtain the apple grade based on fuzzy arithmetic. Finally, we set forth the archetypal non-destructive selection system, which includes robot, AGV, conveyor, sensors and etc. Results of experiments show that this method can distinguish the apple’s quality as well as we set before the experiments, and the non-destructive system is feasible.