نوع مقاله : مقاله پژوهشی
عنوان مقاله English
نویسندگان English
In this paper, we first present a programming-based simulation of an automotive FMCW radar. Then, an automotive radar simulator is designed using Qt software and the C programming language. This simulator initially models targets using the ray tracing method; however, its performance in generating data is subsequently improved by means of real data and a trained model that employs an LSTM network. In the target detection section, the feasibility of utilizing deep learning models proposed for image segmentation has also been investigated. For this purpose, a U‑Net‑based network is first used, which exhibits significantly better performance compared to traditional methods, especially at low SNRs. Finally, the SAM model released by Meta (Facebook) has been tested for this problem and is evaluated as suitable for detecting high‑SNR targets. An improvement of over 70% in detection performance relative to traditional CFAR methods, as well as the capability to identify the target type, are among the advantages of employing these models over CFAR methods.
کلیدواژهها English
