Convolutional Neural Networks for Processing Micro-Doppler Signatures and Range-Azimuth Radar Maps of Frequency Modulated Continuous Wave Radars

• Main Authors: Mardiev, Artem A., Kuptsov, Vladimir D.
• Format: Conference Proceeding
• Language: English
• Subjects: Classification algorithms; convolutional neural network; Image recognition; micro-Doppler signatures; Radar; Radar FMCW; Radar imaging; range-azimuth radar maps; Roads; Target recognition; Training
• ISSN: 2771-697X
• DOI: 10.1109/EExPolytech56308.2022.9950901

The article is devoted to the development of algorithms for radar data processing of continuous-wave radars using convolutional neural networks. The aim of the work is to develop an algorithm for recognizing micro-Doppler signatures based on convolutional neural networks, as well as to develop an algorithm for recognizing range-azimuth radar maps based on convolutional neural networks using video images as training masks. The work was carried out using the Python language, the Tensorflow and Keras machine learning libraries, the Matplotlib library for working with graphs and images. The interactive cloud environment Google Colab was used to train neural networks. The results of the work are the trained convolutional neural networks that are able to effectively recognize and classify unmanned aerial vehicles, cars, cyclists and pedestrians. The developed algorithms can be used in computer vision systems for analyzing radar data. The results of the work show the possibility of using convolutional neural networks in the advanced driver assistance systems of unmanned vehicles.