A novel method combining deep learning with the Kennard-Stone algorithm for training dataset selection for...

Area of Science:

• Agricultural Science
• Computer Science
• Data Science

Background:

• Rice variety identification is crucial due to increased breeding and diverse characteristics.
• Accurate identification impacts trading and agricultural practices.
• Advanced breeding technologies necessitate efficient identification methods.

Purpose of the Study:

• To develop an improved method for rice variety classification.
• To enhance image resolution for better classification accuracy.
• To introduce a novel training sample selection strategy.

Main Methods:

• Collected RGB images of 20 hybrid rice seed varieties.
• Utilized an enhanced deep super-resolution network (EDSR) for image enhancement.
• Integrated deep learning (CNNs, autoencoders) with the Kennard-Stone (KS) algorithm for training sample selection.

Main Results:

• High-resolution images improved variety classification performance.
• The novel training sample selection methodology outperformed random selection by approximately 10.08% in accuracy.
• The proposed methods demonstrated robustness against image noise.

Conclusions:

• Supervised and unsupervised learning models are effective feature extractors.
• Deep learning significantly impacts training set sample selection for classification.
• The study presents a novel approach for training sample selection applicable to various datasets.