A Study on Object Classification Based on Deep Learning Using Radar Signal According to Invisible Area and MIMO Antenna Configuration

The technology for detecting objects in invisible area is a technology for recognizing objects in an area that cannot be seen in general view, and is attracting attention in fields such as military operations, lifesaving, and autonomous driving. The RF radar signal has the characteristic of penetrating wall, so it is considered suitable for detecting an object in a special environment called invisible area. At this time, a loss occurs when the RF signal passes through an obstacle constituting an invisible environment, and the loss results in a lower performance of the invisible object detection technology. In this paper, multiple transmit/receive antenna technology is applied to solve this problem and an RF radar signal collection experiment environment is constructed through an ultra-wideband radar chip. The RF signal dataset is collected while changing the type of the wall and antenna arrangement, and the object classification results output through machine learning and deep learning models were compared and analyzed. Through this, it is verified through whether the object detection performance is affected as the signal loss also varies through various wall and antenna arrangement configurations.