Joint Source-Channel Coding base on DL for Unmanned Aerial Vehicle Image Transmission

Traditional image transmission solutions incorporate source coding to minimize redundant information, addressing bandwidth and power constraints. Following this, channel coding is applied to introduce redundancy, ensuring the efficacy of data transmission. However, unmanned aerial vehicles (UAVs) encounter challenges such as limited signal coverage, poor channel conditions, and constraints on transmission bandwidth and energy during wireless image transmission. The utilization of traditional source-channel concatenated coding for data transmission incurs a high cost. To tackle this challenge, this paper proposes a Joint Source-Channel Coding (JSCC) image transmission solution based on deep learning. This approach directly maps image pixels to complex-valued channels, thereby reducing the overall cost of end-to-end transmission. Furthermore, an optimized JSCC method is introduced, integrating adaptive attention mechanisms, channel estimation, and channel equalization algorithms. Experimental results demonstrate that the proposed method not only outperforms traditional concatenated transmission schemes but also enhances the performance of JSCC in terms of transmission quality.