Deep4D-Radar utilizing Artificial Intelligence and Machine Learning

雷达系统由于其高可靠性和与天气无关的特点，在汽车工业中被广泛用于辅助驾驶和自动驾驶系统。数字信号和图像处理是雷达系统从传感器和采集的数据中检测和分类目标的重要组成部分。通常，调频连续波雷达用于汽车应用。该系统采用恒虚警率算法，经过快速傅立叶变换运算后，在接收端进行目标检测。这些算法依赖于对背景噪声的估计，在多目标情况下给出的结果不准确，从而导致检测精度下降。在与卡尔斯鲁厄应用科学大学的合作中，我们的目标是利用机器学习和人工神经网络从雷达数据中检测和分类目标。利用卷积神经网络(CNN)从雷达信号频谱中提取特征，准确地检测目标，并根据目标的信号模式对其进行分类。其目标是开发能够处理4D雷达信号的傅里叶谱的多维CNN，以提高雷达系统的分辨率和分类能力。在拟议的项目中，申请者预计将利用人工智能和机器学习在数字信号处理的高需求领域培训HQP，这将推动雷达系统新兴研发领域的创新，特别是在汽车行业。该项目将加强我们的国际合作，并将为联盟国际合作赠款申请提供一个平台。此外，所开发的算法将来还可以集成到嵌入式系统中，用于加拿大和德国的各种工业应用。

Radar systems are highly utilized today in automobile industry for driver assistance and autonomous driving systems due to their high reliability and weather independencies. Digital signal and image processing is an integral part of radar systems to detect and classify objects from the sensors and collected data. Typically, frequency modulated continuous wave radar is used for automotive applications. This system employs constant false alarm rate algorithms for object detection in the receiver after a Fast Fourier Transform calculation. These algorithms depend on estimation of background noise, which in the case of a multiple object scenario gives inaccurate results and thus leads to degraded detection accuracy.In this proposed collaboration with Karlsruhe University of Applied Sciences, we aim to detect and classify objects from radar data utilizing machine learning and artificial neural networks. Using convolutional neural networks (CNNs), features will be extracted from the radar signal frequency spectra to accurately detect objects and classify them based on their signal patterns. The goal is to develop multidimensional CNNs capable of processing the Fourier spectra of 4D radar signals to improve the resolution and classification capability of the radar system. In the proposed project, the applicant is expected to train HQPs in high-demand areas of digital signal processing utilizing artificial intelligence and machine learning, which will drive innovation in emerging areas of research and development in radar systems, specifically in the auto industry. This project allows for strengthening our international collaborations and will provide a platform for an Alliance International Collaboration grant application. In addition, the developed algorithms could be integrated into an embedded system in future for various industrial applications in Canada and Germany.