Temporal and Spatial Dynamics of Pine Wilt Disease: Insights from a Hybrid CNN-LSTM Approach

The current research proposes a novel method of multi-classifying the Pine Wilt Disease severity using the hybrid model that combines Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks. The paper aims to address the most critical issue of accurately distinguishing between five levels of disease severity, and it implements CNNs to extract spatial features and LSTMs to process temporal data. Training and testing over a complete dataset consisting of images of pine trees affected by Pine Wilt Disease is rigorously performed to maintain the robustness and accuracy of the model in real-world usage. The methodology involves an extensive image preprocessing phase that includes normalization and augmentation, thereby ensuring that the model is exposed to a large set of disease manifestations. With CNNs for detailed feature extraction and LSTMs for capturing the disease's progression over time as the base, the hybrid model shows a sophisticated understanding of the Pine Wilt Disease severity, while exceeding the traditional diagnostic methods' limits. The evaluation of the model confirms an overall accuracy of 98.13% in the classification of disease severity into the set of five levels, which constitutes a substantial improvement over existing models. This accuracy substantiates the model as a good resource for forest management which in turn will help in efficient monitoring, pest control, and mitigation of Pine Wilt Disease. Moreover, the comparative analysis brings forward the improved ability of the CNN-LSTM hybrid model to take charge of temporal aspects, resulting in a complete solution for the complex classification problem. Thus, not only does this study show the practicability and efficiency of consulting a hybrid deep learning method for plant disease severity classification, but it also provides new prospects for plant pathology and disease management in the future. The results thus go towards the wider domain of agricultural technology giving suggestions that may be used for the detection and classification of different plant diseases and hence favoring sustainable forestry practices and ecosystem preservation.