UAV Image Transmission Algorithm Based on Convolutional Neural Networks Under Complex Disaster Conditions
Author:Hailong Yang, Shuai Li, Qiuhong Hu, Chunxue Wu Author Unit: Aerospace Engineering University, Beijing, China Aerospace Dongfanghong Satellite Co., Ltd. School of Optical-Electrical and Computer Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China
Abstract：In recent years, the UAV image transmission method based on neural network structure has been widely concerned. Although the traditional transmission mode has a high speed, it has a series of problems, such as a large amount of calculation, affected by various channel noise, and so on, which cannot fully adapt to the complex disaster environment. In this paper, a method of image transmission combining deep learning and compression sampling is proposed. Only a small amount of data needs to be sampled to solve the problems of compression, transmission and reconstruction at both ends of the system, so as to ensure high-quality data transmission. Compared with the traditional compressed sensing transmission method, this paper adopts a more efficient method. In the reconstruction, many iterations are replaced by a large number of neural network training to achieve the real-time reconstruction and the improvement of image quality, which is conducive to the practical application of compressed sensing technology in the field of complex disaster image processing.
