Drone Measured Tree Species Dataset

Abstract\u2014Remote sensing scene image classiffcation is a critical task in remote sensing interpretation, playing a vital role in enhancing image understanding capabilities. Although deep learning-based methods have been widely applied and signiffcant progress has been made by leveraging multi-source data from unmanned aerial vehicles for feature extraction, challenges such as limited spatial resolution of small-sized images, image blur, and data sparsity continue to constrain model performance. Therefore, there is an urgent need to design more robust and accurate deep learning models to achieve precise classiffcation and efffcient utilization of remote sensing images.To address the issue of insufffcient accuracy in tree species identiffcation from remote sensing images, this study proposes a lightweight novel attentionbased classiffcation network named Treeformer. The model integrates an FFN feature enhancement module, a CMHA multihead attention mechanism, and a gated three-branch attention module. Additionally, it incorporates a multi-scale hybrid convolution structure combining grouped convolution and depthwise separable convolution to enhance feature extraction and suppress redundant information. The introduction of the three-branch attention mechanism, which combines multi-head attention and Fourier transform, efffciently processes complex tree structure data. Moreover, through a modular stacking architecture and frequency-domain information guidance, the model effectively improves its discriminative capability and global perception of complex tree species images. Experimental results demonstrate that the proposed method outperforms mainstream deep learning models across multiple evaluation metrics, achieving a highest classiffcation accuracy of 98.55%.In summary, this study constructs the Treeformer model by integrating multi-scale hybrid convolution and three-branch attention mechanisms, validating its effectiveness in improving accuracy and discriminative ability in tree species image classiffcation.