Research on Identification Method of Red Date Grade Based on Deep Learning

Youjiang Yu

《International Core Journal of Engineering》 7卷12期 (2021/12) Pp. 11-17

https://doi.org/10.6919/ICJE.202112_7(12).0002

摘要

Red dates need to be graded in the process of processing, so as to price according to quality, convenient for consumers to choose at the same time to speed up the circulation of goods, improve the added value of products. A deep learning based Stacking method for date classification was proposed. 21280 date images of 4 different grades of the same variety were collected for pre-processing and data set was established. A Stacking integrated learning model with three different convolutional neural networks (VGG16, ResNet 50, Densenet121) as the base learner and logistic regression as the secondary learner was constructed. The comparison between the integrated model and the single neural network model was conducted. The results show that the Stacking integration model proposed in this experiment can achieve 91.88% accuracy in jujube grading.

關鍵字

Deep learning ； Classification ； Red jujube ； Convolutional neural network ； base learner

參考文獻

Li L, Wang Y. Improved LeNet-5 Convolutional Neural Network Traffic Sign Recognition[J]. International Core Journal of Engineering, 2021, 7(4).

Yu W, Dong-Fei W, Hui-Feng W, et al. A microfluidic robot for rare cell sorting based on machine vision identification and multi-step sorting strategy [J]. Talanta, 2021,226.

Mon T, ZarAung N. Vision based volume estimation method for automatic mango grading system [J]. Biosystems Engineering, 2020, 198.

Cárdenas-Pérez S, Chanona-Pérez J, Méndez-Méndez J V, et al. Evaluation of the ripening stages of apple (Golden Delicious) by means of computer vision system [J]. Biosystems Engineering, 2017, 159.

Traffano-Schiffo M V, Castro-Giraldez M, Colom R J, et al. New Spectrophotometric System to Segregate Tissues in Mandarin Fruit [J]. Food and Bioprocess Technology, 2018, 11 (2).