High-throughput corn ear screening method based on two-pathway convolutional neural network
2020
Zhang, JiaJing | Ma, Qin | Cui, Xuelian | Guo, Hao | Wang, Ke | Zhu, DeHai
Screening corn ear is an important and time-consuming task in the process of corn seed production. Recently, Deep Learning plays a major role in learning image features, which has widely used in image classification tasks. In this paper, we propose a novel high-throughput method to screen corn ear at actual seed production factory based on the two-pathway convolution neural network. The two-pathway convolution neural network combined the advantages of VGG-16 and Resnet-50 can greatly reduce input parameters and improve accuracy. Training of the network was performed with the use of an enlarged corn ears dataset, which was acquired by our high-throughput and easy-to-clean image acquisition device. The dataset contains 17,502 corn images, including 6320 normal corn, 5422 rehusk corn and 5760 abnormal corn. The experiments showed that the average classification accuracy was 97.23%. To ensure the superior performance of the proposed two-pathway convolution neural network, we have carried out the experimental comparison with other one-way networks. Also, we explore the impact of network parameters on the classification results. Comprehensive empirical analyses reveal that the proposed method achieves excellent performance and outperforms existing methods in a non-structural environment.
Показать больше [+] Меньше [-]Ключевые слова АГРОВОК
Библиографическая информация
Эту запись предоставил National Agricultural Library