The evolution of the sixth-generation (6G) communication technology has brought new opportunities for the development of the integrated sensing and communication (ISAC). Recently, with the growth of low-altitude digital economy, the identification of the commercial unmanned aerial vehicle (UAV) has become an important application scenario for the ISAC. Based on deep learning, this paper proposes an intelligent identification method for the commercial UAV by using an improved residual network (ResNet), namely wireless-ResNet. Specifically, considering the typical wireless challenges of the UAV radio frequency (RF) identification such as the non-stationary characteristics of the UAV signals and the response speed of the UAV identification, we propose a wireless-ResNet by optimizing the data preprocessing, the model structure and parameters. The identification performance of the UAV signals is significantly improved under the real-world RF datesets of the commercial UAV. The experimental results demonstrate that the proposed method excels in both identification accuracy and response speed, achieving an identification accuracy of 0.92 and an average response time of 18.52 ms, respectively, which outperforms the classic ResNet and the convolution neural network (CNN) models.