Cross-Sectional Texture and Pattern Analysis of Burmese and Assamese Areca Nuts Using CNN Model

This paper endeavours to distinguish between Burmese areca nut and Assamese areca nut based on their cross-sectional texture and pattern. Employing advanced machine learning techniques, specifically Convolutional Neural Networks (CNNs), we constructed a model capable of accurately classifying these two variants using images of their cross-sections. The paper explores two distinct datasets: one comprising unedited images and the other generated by applying edge detection algorithms to these images. A custom Python script utilizing OpenCV libraries was developed to create edge maps, providing enhanced feature representations for the CNN model. The model architecture involves convolutional layers, max-pooling layers, and dropout layers for regularization. We conducted extensive experimentation with varied batch sizes and epochs to optimize the model's performance. Our results reveal that the model achieved high accuracy in distinguishing between the two areca nut variants, demonstrating the potential for machine learning to contribute to quality control and identification in this domain. This paper not only presents a successful classification model but also sheds light on the efficacy of edge maps as input data for image-based classification tasks.