Lightweight Semantic Communication Model Driven UAV for Intelligent Transmission

Unmanned Aerial Vehicles (UAVs) are crucial tools for remote sensing tasks. However, UAVs face limitations in computational and communication resources. In the context of the exponential growth of remote sensing data, it is necessary that this captured information be transmitted to powerful ground servers for better processing. To accomplish this task efficiently, this paper proposes an intelligent transmission scheme based on semantic communication. An end-to-end encoder-decoder network is trained to understand the semantic features of remote sensing images, ultimately reconstructing high-quality original images, which significantly reduces channel resource overhead. Additionally, to further reduce the computational load on the transmitter to better align with practical constraints, this paper introduces a pruning method based on $l_{1}$-norm regularization, which reduce the overall computational cost of the model while maintaining performance. Simulation results demonstrate that, compared to baseline models, the proposed system features a more lightweight architecture and superior performance, even under low signal-to-noise ratio (SNR) conditions.