Objective To solve the problems such as poor grading accuracy and low efficiency existing in the current automatic grading methods for apples.Methods On the basis of the automatic grading system for apples based on machine vision, an automatic grading method combining convolutional neural network, global average pooling, batch normalization, and support vector machine is proposed for apples. Through global average pooling, the number of model parameters is reduced. The generalization ability of the model is improved by batch normalization. The Softmax classifier of the convolutional neural network is replaced by a support vector machine to improve the grading accuracy. Finally, verification tests are carried out.Results Compared with conventional grading methods for apples, the automatic grading method established in this study has increased accuracy and efficiency, with the grading accuracy of 98.50% and the grading speed of 209 FPS, which meets the requirements of food processing automation.Conclusion By optimizing the existing automatic grading methods for apples, the detection performance is improved to a certain extent.