Deepening the Accuracy of Tree Species Classification: A Deep Learning-Based Methodology

Abstract

The utilization of multi-temporally integrated imageries, combined with advanced techniques such as convolutional neural networks (CNNs), has shown significant potential in enhancing the accuracy and efficiency of tree species classification models. In this study, we explore the application of CNNs for tree species classification using multi-temporally integrated imageries. By leveraging the temporal variations captured in the imageries, our goal is to improve the classification models’ discriminative power and overall performance. The results of our study reveal a notable improvement in classification accuracy compared to previous approaches. Specifically, when compared to the random forest model’s classification accuracy of 84.5% in the Gwangneung region, our CNN-based model achieved a higher accuracy of 90.5%, demonstrating a 6% improvement. Furthermore, by extending the same model to the Chuncheon region, we observed a further enhancement in accuracy, reaching 92.1%. While additional validation is necessary, these findings suggest that the proposed model can be applied beyond a single region, demonstrating its potential for a broader applicability. Our experimental results confirm the effectiveness of the deep learning approach in achieving a high accuracy in tree species classification. The integration of multi-temporally integrated imageries with a deep learning algorithm presents a promising avenue for advancing tree species classification, contributing to improved forest management, conservation, and monitoring in the context of a climate change.