Classification of Corn Seed Quality Using Convolutional Neural Network with Region Proposal and Data Augmentation

Satoto, Budi Dwi and Wahyuningrum, Rima Tri and Khotimah, Bain Khusnul (2023) Classification of Corn Seed Quality Using Convolutional Neural Network with Region Proposal and Data Augmentation. Jurnal Ilmiah Teknik Elektro Komputer dan Informatika (JITEKI), 9 (2). pp. 348-362.

Text 11-Classification of Corn Seed Quality Using Convolutional Neural Network with Region Proposal and Data Augmentation.pdf Download (1MB)

Abstract

Corn is a commodity in agriculture and essential to human food and animal feed. All components of corn can be utilized and accommodated for the benefit of humans. One of the supporting components is the quality of corn seeds, where specific sources have physiological properties to survive. The problem is how to get information on the quality of corn seeds at agricultural locations and get information through direct visual observations. This research tries to find a solution for classifying corn kernels with high accuracy using a convolutional neural network. It is because in-depth training is used in deep learning. The problem with convolutional neural networks is that the training process takes a long time, depending on the number of layers in the architecture. The research contribution is adding Convex Hull. This method looks for edge points on an object and forms a polygon that encloses that point. It helps increase focus on the convolution multiplication process by removing images on the background. The 34-layer architecture maintains feature maps and uses dropout layers to save computation time. The dataset used is primary data. There are six classes, AR21, Pioner_P35, BISI_18, NK212, Pertiwi, and Betras1—data augmentation techniques to overcome data limitations so that overfitting does not occur. The results of the classification of corn kernels obtained a model with an average accuracy of 99.33%, 99.33% precision, 99.33% recall, and 99.36% F-1 score. The computational training time to obtain the model was 2 minutes 30 seconds. The average error value for MSE is 0.0125, RMSE is 0.118, and MAE is 0.0108. The experimental data testing process has an accuracy ranging from 77% -99%. In conclusion, using the proposal area can improve accuracy by about 0.3% because the focused object helps the convolution process.

Item Type:	Artikel Umum
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering
Divisi / Prodi:	Faculty of Industrial Technology (Fakultas Teknologi Industri) > S1-Electrical Engineering (S1-Teknik Elektro)
Depositing User:	M.Eng. Alfian Ma'arif
Date Deposited:	29 May 2023 01:19
Last Modified:	29 May 2023 01:19
URI:	http://eprints.uad.ac.id/id/eprint/43218

Actions (login required)

View Item