Research on radar clutter recognition method based on LSTM

Abstract

Considering the increasingly complex electromagnetic environment of radar detection, it still contains a lot of clutter after target detection, which is a challenging problem for subsequent target tracking. A radar clutter recognition method based on long short-term memory (LSTM) neural network is proposed here. This method can further identify clutter points after target detection and improve the quality of target tracking. LSTM neural network is trained and compared with K nearest neighbour method and support vector machine. It is found that the clutter recognition accuracy of this method is up to 86.3%, which is 12.2 and 4.1% higher than the latter two methods. At the same time, the target loss rate of this method is only 16.6%, which is 18.3 and 5.7% lower than the latter two methods respectively. The experimental results show that the radar clutter recognition method based on LSTM neural network is effective.

Keywords

• radar clutter
• target tracking
• object detection
• radar detection
• radar signal processing
• recurrent neural nets
• learning (artificial intelligence)
• short-term memory neural network
• clutter points
• target detection
• LSTM neural network
• K nearest neighbour method
• clutter recognition accuracy
• radar clutter recognition method
• radar detection
• complex electromagnetic environment
• target tracking
• support vector machine