Enhancing Weed Detection with Convolutional Nerual Networks for Precision Farming

Le résumé fourni par la source

The aim of this research is to compare the effectiveness of a Convolutional Neural Network (CNN)-based model with a Support Vector Machine (SVM)-based method in order to improve weed detection for precision farming. Materials and Methods: Materials and Methods: Two groups were assessed for this study. A CNN-based weed detection model used in Group 1 was trained and evaluated on 26 samples of agricultural field photos that included a range of environmental factors, including different lighting, overlapping vegetation, and uneven field layouts. Before training, image processing methods such as color, texture, and edge enhancements were used. The same 26 image samples were used to train and test Group 2's SVM-based weed detection model, and features were manually extracted using Local Binary Patterns (LBP) and Histogram of Oriented Gradients (HOG). The threshold is set at 0.05% with a 95% confidence interval, and the G Power value is set at 80%. Result: The accuracy of the CNN-based model in Group 1 was 93%, which was significantly higher than that of the SVM-based model in Group 2, which was 81%. Furthermore, the CNN-based model demonstrated improved robustness in challenging field conditions and quicker processing times. Conclusion: When it comes to accuracy, speed, and managing challenging field conditions, the CNN-based weed detection model outperforms the SVM-based model by a wide margin. Real-time, scalable applications are made possible by its integration with IoT devices, which also helps to promote sustainable farming and lessen the use of herbicides.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.