Details

Autor(en) / Beteiligte

• Ren, Wenchao
• Liu, Zeyuan
• Lv, Rongxin
• Ge, Yilin
• Yen, Neil
• Chang, Jia-Wei
• Hung, Jason C
• Hung, Jason C.

Titel

Research on Radar Target Recognition Based on Deep Learning

Ist Teil von

• Frontier Computing on Industrial Applications Volume 2, 2024, Vol.1132, p.108-114

Ort / Verlag

Singapore: Springer Singapore Pte. Limited

Erscheinungsjahr

2024

Quelle

Alma/SFX Local Collection

Beschreibungen/Notizen

• Traditional shallow learning radar target image recognition ways usually depend on complex artificial feature extraction process, which needs a lot of human expertise. to effectively extract the characteristic information of enemy combat units, and realize the type recognition of the UAV radar, which is very important to ensure the UAV’s combat capability and combat victory. Aiming at the problem that the shallow learning method of radar target image recognition is difficult to extract advanced features and the small number of enemy unit radar image samples available in practical engineering has a great effect on the diagnosis accuracy of the deep neural network model, this paper uses the transfer method in deep learning. The learning way is to fine-tune the pre-trained hotspot deep CNN convolutional network (GoogleNet) respectively, which are used for the recognition and classification of radar target images. The noise reduction radar target figure is used as input to train a constructed model, Then, the trained model can realize radar target image recognition and classification. The model are validated using the MSTAR dataset gived by DARPA/AFRL, and the results show that model achieve high diagnostic accuracy, proving the effectiveness of the proposed way in radar image recognition of enemy combat units.