Fast and Accurate Classification of Corn Varieties Using Deep Learning With Edge Detection Techniques | JoVE Visualize

Area of Science:

Agricultural Science
Computer Science
Machine Learning

Background:

Accurate corn grading is crucial for maintaining product quality, optimizing storage, and reducing agricultural losses.
Traditional classification methods struggle with large datasets, necessitating faster and more accurate approaches.
Deep learning offers a promising avenue for automated and efficient agricultural product classification.

Purpose of the Study:

To explore a faster and accurate method for classifying corn varieties using deep learning models.
To evaluate the performance of ResCNN, DAG-Net, and ResNet-18 models in classifying three distinct corn varieties.
To compare the effectiveness of different image preprocessing techniques (Canny edge detection, Sobel edge detection, and normal color images) on classification accuracy.

Main Methods:

Three deep learning models (ResCNN, DAG-Net, ResNet-18) were employed for classification.
A dataset of 1050 corn images, representing Chulpi Cancha, Indurata, and Rugosa varieties, was utilized.
Images were preprocessed using Canny edge detection algorithm (CEDA), Sobel edge detection algorithm (SEDA), and normal color images (CI) to create three distinct datasets.

Main Results:

All tested deep learning models achieved high classification accuracies, often exceeding 99% across different datasets and corn varieties.
ResCNN, DAG-Net, and ResNet-18 models demonstrated faster training times compared to using normal color images (CI).
Specific accuracy values varied slightly depending on the image preprocessing method and the deep learning model used, with near-perfect scores for some varieties.

Conclusions:

Deep learning models, including ResCNN, DAG-Net, and ResNet-18, are highly effective for rapid and accurate corn variety classification.
Image preprocessing techniques like CEDA and SEDA can be successfully integrated with deep learning for enhanced corn image analysis.
The study highlights the potential of deep learning to significantly improve the efficiency and sustainability of corn grading processes in agriculture.