Soni Karan, Chandra Gangwar Rakesh
Department of Computer Science Engineering, Sardar Beant Singh State University, Gurdaspur, Punjab, India.
Planta. 2025 Aug 14;262(4):81. doi: 10.1007/s00425-025-04797-9.
This survey concludes that CNN-based deep learning models offer opportunities for early and accurate plant disease detection, supporting sustainable agriculture while acknowledging potential challenges in practical real-world application. Deep learning (DL) methods have transformed image-based plant disease diagnosis by addressing complex challenges specific to crop health monitoring in agriculture. The automated identification and classification of plant disease from images have significant interest, which can be expected to increase crop health monitoring and agricultural productivity. Yet, notwithstanding these benefits, image-based identification of plant diseases is a sophisticated challenge. Proper identification of certain plant varieties and proper determination of disease manifestations are key factors in the administration of effective care and sustainable management of disease. In this research paper an extensive overview Convolutional Neural Networks (CNNs) is implemented using deep learning method for disease detection in plants. This article focuses particularly on highlighting recent research achievements by the last half-decade emphasizing CNN-based models constructed for detecting plant leaf disease. The survey delves into key innovations, methods, and issues faced with the application of CNNs to monitor plant health. Specifically, it highlights the manner in which deep convolutional neural networks (DCNNs), learned using large-scale image databases, are becoming effective means of early and precise detection of plant diseases. Lastly, this paper charts exciting future directions for DL-aided plant disease diagnosis, while providing a balanced critique of the potential, as well as the limitations of CNNs in practical agricultural contexts.
本次调查得出结论，基于卷积神经网络（CNN）的深度学习模型为植物病害的早期准确检测提供了机会，在支持可持续农业的同时，也认识到实际应用中可能面临的挑战。深度学习（DL）方法通过应对农业作物健康监测中特有的复杂挑战，改变了基于图像的植物病害诊断。从图像中自动识别和分类植物病害备受关注，有望加强作物健康监测并提高农业生产力。然而，尽管有这些好处，基于图像的植物病害识别仍是一项复杂的挑战。正确识别某些植物品种以及准确确定病害表现，是有效护理和病害可持续管理的关键因素。在本研究论文中，使用深度学习方法对卷积神经网络（CNNs）进行了广泛概述，以用于植物病害检测。本文特别着重强调过去五年的最新研究成果，重点是为检测植物叶片病害构建的基于CNN的模型。该调查深入探讨了将CNNs应用于监测植物健康所面临的关键创新、方法和问题。具体而言，它突出了通过大规模图像数据库学习的深度卷积神经网络（DCNNs）如何成为早期精确检测植物病害的有效手段。最后，本文规划了深度学习辅助植物病害诊断令人兴奋的未来方向，同时对CNN在实际农业环境中的潜力和局限性进行了全面的审视。
