High-Resolution Range Profile Recognition Method of Vehicle Targets Based on Accelerated T-SNE with Multi-polarization Fusion
Hao Wu,Dahai Dai,Penghui Ji,Bo Pang,Xuesong Wang
DOI: https://doi.org/10.1109/icsip52628.2021.9688626
2021-01-01
Abstract:Polarization plays an important role in analyzing radar targets. Full polarization information has a better ability to inverse fine structures of the target. However, previous studies of high-resolution range profile (HRRP) mainly neglected effects of polarization. Therefore, a novel radar target recognition algorithm based on accelerated t-Distributed Stochastic Neighbor Embedding(t-SNE) and multi-polarization weighted fusion has been proposed. Firstly, three single polarization channels, odd and even-bounce mechanism HRRP, and full polarization HRRP are described. The accelerated t-SNE method based on Barnes-Hut approximation improves the dimensionality reduction speed of processing HRRPs while maintaining the initial information of each HRRP. Also, it is capable of visualizing different targets. With the weighted fusion of posterior probabilities of multiple support vector machines (SVMs), we are able to realize recognition of targets. Experimental results of vehicle targets based on electromagnetic calculations show that recognition performance of this algorithm is superior to other methods, and it shows strong robustness under the condition of large azimuth angle.