This study presents an automated approach for detecting Tomato Leaf Blight using convolutional neural networks (CNNs) applied to leaf images. The methodology leverages image preprocessing techniques, such as segmentation and feature extraction, to enhance classification accuracy. A dataset of 5,000 tomato leaf images was used to train and validate the CNN model, achieving an accuracy of 92%. The results demonstrate the potential of CNNs in early disease detection, aiding farmers in timely interventions. Future work includes integrating real-time monitoring systems.
Introduction
Tomato Leaf Blight, caused by Phytophthora infestans, is a major threat to tomato crops worldwide. Early and accurate detection is essential to reduce losses, but manual inspection is inefficient. Advances in image processing and machine learning, especially convolutional neural networks (CNNs), have greatly improved automated plant disease detection by effectively extracting complex features from leaf images.
This study developed a CNN-based model to classify healthy and diseased tomato leaves using a dataset of 5,000 images (half healthy, half diseased). Images were preprocessed through resizing, normalization, and K-means clustering segmentation to isolate leaves and reduce noise. Features such as texture (via GLCM) and color (in HSV space) were extracted to enhance model performance. The CNN architecture included multiple convolutional and pooling layers, trained with the Adam optimizer.
The model achieved 92% accuracy, outperforming traditional methods like SVM and Random Forest. K-means segmentation improved accuracy by 5%. The model showed strong precision (0.91), recall (0.93), and an AUC of 0.95, indicating robust classification. Challenges noted included overfitting and limited dataset diversity, which could be mitigated by data augmentation.
Conclusion
Future enhancements to this study involve integrating the proposed model into a mobile application for real-time disease detection, thereby increasing its accessibility and practical impact in agricultural settings. Additionally, incorporating multi-spectral imaging can enable more comprehensive feature extraction, capturing data beyond the visible spectrum to improve diagnostic accuracy (Li et al., 2020). Expanding the dataset to include a broader range of tomato diseases will enhance the model’s generalizability and robustness across different disease types. Furthermore, applying transfer learning using pre-trained deep learning architectures such as ResNet could significantly improve classification performance by leveraging learned features from large-scale image datasets (Tan et al., 2018).
This study demonstrates the effectiveness of combining convolutional neural networks (CNNs) with image processing techniques for the detection of Tomato Leaf Blight, achieving an accuracy of 92%. The methodology presents a scalable and automated approach for early disease identification, which is crucial for timely intervention in crop management. With ongoing advancements in deep learning algorithms and imaging technologies, such approaches hold substantial promise for enhancing precision agriculture and supporting sustainable farming practices.
References
[1] Arivazhagan, S., et al. (2013). Detection of unhealthy region of plant leaves using image processing. Computers and Electronics in Agriculture, 92, 74-82.
[2] Barbedo, J. G. A. (2016). A review on the use of image processing for plant disease detection. Computers and Electronics in Agriculture, 129, 156-163.
[3] Brahimi, M., et al. (2017). Deep learning for tomato disease classification. Procedia Computer Science, 112, 279-286.
[4] Ferentinos, K. P. (2018). Deep learning models for plant disease detection. Computers and Electronics in Agriculture, 145, 311-318.
[5] Fry, W. E. (2008). Phytophthorainfestans: The plant destroyer. Molecular Plant Pathology, 9(3), 385-402.
[6] Hughes, D., &Salathé, M. (2015). An open access repository of plant images. PlantVillage Dataset.
[7] Kamilaris, A., &Prenafeta-Boldú, F. X. (2018). Deep learning in agriculture: A survey. Computers and Electronics in Agriculture, 147, 70-90.
[8] Li, L., et al. (2020). Multispectral imaging for plant disease detection. Remote Sensing, 12(4), 614.
[9] Luque, A., et al. (2018). Evaluation of ROC curves in plant disease classification. Pattern Recognition Letters, 113, 45-52.
[10] Mohanty, S. P., et al. (2016). Using deep learning for image-based plant disease detection. Frontiers in Plant Science, 7, 1419.
[11] Polder, G., et al. (2014). Image segmentation for plant phenotyping. Plant Methods, 10(1), 23.
[12] Saleem, M. H., et al. (2019). Plant disease detection using machine learning: A review. Biosystems Engineering, 187, 75-89.
[13] Shorten, C., &Khoshgoftaar, T. M. (2019). A survey on image data augmentation for deep learning. Journal of Big Data, 6(1), 60.
[14] Singh, V., et al. (2019). Random forest for plant disease classification. Precision Agriculture, 20(4), 767-781.
[15] Sladojevic, S., et al. (2016). Deep neural networks for plant disease recognition. Computational Intelligence and Neuroscience, 2016, 1-10.
[16] Tan, C., et al. (2018). Transfer learning for plant disease detection using ResNet. IEEE Access, 6, 42587-42596.
[17] Too, E. C., et al. (2019). A comparative study of CNN architectures for plant disease detection. Agriculture, 9(6), 119.
[18] Wang, G., et al. (2017). Deep learning for plant disease detection: A survey. IEEE Transactions on AgriFood Electronics, 5(3), 123-135.
[19] Zhang, K., et al. (2019). Image preprocessing for plant disease detection. Journal of Imaging, 5(4), 45.
[20] Zhang, X., et al. (2018). SVM-based plant disease classification. Computers and Electronics in Agriculture, 152, 185-192.
[21] https://www.vegarden.com/2010/07/tomato-blight/