Vol.27, No.2, 2021, pp.581-592, doi:10.32604/iasc.2021.014152
OPEN ACCESS
ARTICLE
Study of Sugarcane Buds Classification Based on Convolutional Neural Networks
  • Huaning Song1, Jiansheng Peng1,2,*, Nianyang Tuo1, Haiying Xia2, Yiyun Peng3
1 Hechi University, Yizhou, 546300, China
2 Guangxi Normal University, Guilin, 541004, China
3 JPMorgan Chase & Co. 270 Park Avenue, New York,10017, USA
* Corresponding Author: Jiansheng Peng. Email:
Received 02 September 2020; Accepted 30 October 2020; Issue published 18 January 2021
Abstract
Accurate identification of sugarcane buds, as one of the key technologies in sugarcane plantation, becomes very necessary for its mechanized and intelligent advancements. In the traditional methods of sugarcane bud recognition, a significant amount of algorithms work goes into the location and recognition of sugarcane bud images. A Convolutional Neural Network (CNN) for classifying the bud conditions is proposed in this paper. Firstly, we convert the colorful sugarcane images into gray ones, unify the size of them, and make a TFRecord format data set, which contains 1100 positive samples and 1100 negative samples. Then, a CNN mode is developed to classify the buds into good ones or bad ones. This model contains two convolutional layers and two pooling layers, which go into two full connected layers at the end. The bud dataset is divided into three sets: training set (80%), validation set (10%) and test set (10%). After the repeated training, the structure and parameters of the model are optimized, and an optimal application model is obtained. The experiment results demonstrate that the recognition accuracy of sugarcane bud reaches to 99%, which is about 6% higher than traditional methods. It proves that the CNN model is feasible to classify sugarcane buds.
Keywords
Sugarcane bud; classification; convolutional neural networks
Cite This Article
H. Song, J. Peng, N. Tuo, H. Xia and Y. Peng, "Study of sugarcane buds classification based on convolutional neural networks," Intelligent Automation & Soft Computing, vol. 27, no.2, pp. 581–592, 2021.
This work is licensed under a Creative Commons Attribution 4.0 International License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.