Abstract
The traditional SAR motion parameter estimation method based on physical or statistical models has problems such as poor versatility and low efficiency. Starting from SAR defocused images, combined with deep learning technology is expected to solve the above problems. To this end, this paper studies a SAR motion error parameter estimation algorithm combined with the convolutional neural network, which uses GoogLeNet network architecture to build training networks. Based on the idea of image classification, the motion error of the airborne SAR is estimated. Finally, the influence of scene types on parameter estimation network training is analyzed experimentally.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have