A Novel Deep Convolutional Classifier Model for NDVI Image Recognition

Forests are a vital component of terrestrial ecosystems. Investigating the spatiotemporal characteristics of regional forest cover changes and predicting future forest conditions can provide crucial guidance for forest resource conservation and sustainable ecological development. This study proposes a Deep Convolutional Classifier Model (DCCM) for forest area prediction and explores the spatiotemporal evolution of forest and non-forest regions in central and western China from 2003 to 2023. The results indicate: (1) The mean NDVI values in central and western China were <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$0.54,0.59$</tex>, and 0.59 in 2003, 2013, and 2023, respectively, demonstrating an overall increasing trend in vegetation coverage with strong regional geographical heterogeneity and significant variations in vegetation distribution; (2) Image segmentation results reveal a notable expansion in the proportion of forest areas with NDVI greater than 0.7 during 2003-2023; (3) The DCCM's performance in predicting segmentation outcomes, as measured by four metrics-Accuracy, Precision, Recall, and F1—Scoreoutperforms comparative models in comprehensive evaluation, with predictive capability improving annually, indicating the model's exceptional generalization ability and effectiveness in capturing forest features. This research provides a scientific foundation for achieving forest conservation and sustainable development.