A Novel Recognition Method of ISAR Image Fusion Based on Botnet and Embracenet
Authors:
- 22 October 2021
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 841-845
Abstract
Compared with single ISAR image recognition, multiple ISAR images contain more target information, which is more conducive to target recognition. Simultaneously considering the non-cooperation and complexity of target motion, ISAR image usually has low-resolution or even cannot be generated, this paper proposes a feature fusion model for ISAR images which takes account of both recognition rate and robustness. In this model, feature information is extracted based on Botnet and multi-sensor features are fused based on Embracenet. The experimental results based on the simulation data set show that in the five types of target recognition tasks under different elevation angles, the classification accuracy of the proposed method based on the fusion features of two ISAR images is better than that of single ISAR image. In the absence of one ISAR image, the recognition rate decreased slightly, and was still higher than that of the single ISAR image method.Keywords
This publication has 4 references indexed in Scilit:
- Bottleneck Transformers for Visual Recognition2021
- High-Performance SAR Automatic Target Recognition Under Limited Data Condition Based on a Deep Feature Fusion NetworkIEEE Access, 2019
- EmbraceNet: A robust deep learning architecture for multimodal classificationInformation Fusion, 2019
- Multi-Stream Convolutional Neural Network for SAR Automatic Target RecognitionRemote Sensing, 2018