A Machine Learning Perspective on Daisy and Dandelion Classification: Gaussian Naive Bayes with Sobel

Abstract

This study explores the classification of Daisy and Dandelion flowers using a Gaussian Naive Bayes classifier, enhanced by Sobel segmentation and Hu moment feature extraction. The research adopted a quantitative approach, utilizing a balanced dataset of Daisy and Dandelion images. The Sobel operator was employed for image segmentation, accentuating the floral features crucial for classification. Hu moments, known for their invariance to image transformations, were extracted as features. The Gaussian Naive Bayes algorithm was then applied, with its performance evaluated through a 5-fold cross-validation process. The results exhibited moderate accuracy, with the highest recorded at 60%, and precision peaking at 62.60%. These findings indicate a reasonable level of effectiveness in distinguishing between the two species, though variations in performance metrics suggested room for improvement. The study contributes to the field of botanical image classification by demonstrating the potential of integrating image processing techniques with machine learning for flower classification. However, it also highlights the limitations of the Gaussian Naive Bayes approach in handling complex image data. Future research directions include exploring more advanced machine learning algorithms and expanding the feature set to enhance classification accuracy. The practical implications of this research extend to ecological monitoring and agricultural studies, where efficient and accurate plant classification is vital