Abstract: Efficient measurement of tree canopy cover in urban environments is important for environmental regulation and urban governance. In this paper, we propose a method that combines deep learning and single-head self-attention mechanism (SHSA) to automatically compute urban tree canopy coverage via a canopy segmentation algorithm. High-resolution RGB images were obtained using UAV aerial photography to construct the dataset. Two segmentation algorithms, YOLOv8 and YOLOv11, were improved, and the coverage area was calculated according to the segmentation results and most compared with the real coverage area of the tree canopy. The experimental data show that YOLOv8 has better segmentation ability compared with YOLOv11, but the training time is longer than YOLOv11. After equipped with this single-head self-attention mechanism, the accuracy of both algorithms is improved, and the training time is significantly reduced, and the accuracy of YOLOv8 is improved by an average of 0.01346, and with the increase of the complexity of the weights, the more significant the reduction of the training time can be. Meanwhile, YOLOv8 also has superior results in the inference of high-resolution RGB images, which is also better than YOLOv11.