A Study on 3D Object Detection in Invisible Area Using Radar Signal and Machine Learning

Technology for detecting objects in invisible spaces is attracting attention for various purposes such as military, lifesaving, and autonomous driving. RF radar signals are considered as suitable sensor types for performing this task because they can measure objects through walls. In this paper, a data collection experiment environment is constructed through a MIMO(Multi-In-Multi-Out) antenna and a Ultra-Wideband radar chip. Using signals collected using the configured environment as datasets and the corresponding dataset as input of the Transformer model, 3D object detection through Bird-Eye-View Bounding Box is performed to present algorithms for object position estimation.