Feature Extraction Method of UAV Life Detection Radar Signal Based on Markov Transition Field

Considering the high complexity of the UAV life detection radar signal and the large amount of data, it is difficult to distinguish the detection target quickly and intelligently. This paper proposes an identification and classification method for life detection radar based on Markov Transition Field and convolutional neural network (CNN). In this paper, the radar echo is down-sampled and processed, and then the one-dimensional data is converted into two-dimensional images by Markov Transition Field. Finally, the feature extraction of the two-dimensional image is performed through the CNN. The analysis of the measured data shows that the average recognition accuracy of the proposed model reaches 95%. It is better than spectrum graph processing and other time series processing methods such as GAF (Gramian Angular Field). It is based on the research on human identification methods of life detection radar provides the basis.