An FMCW MIMO Radar-Vision Fusion Algorithm for Target Classification and Localization

Abstract

In order to solve the problems of slow real-time and poor detection of small targets that still exist in multi-sensor information fusion technology, a frequency-modulated continuous wave (FMCW) multiple-input multiple-output (MIMO) radar-vision fusion algorithm for target classification and localization is proposed in this paper. Firstly, the signal model of FMCW MIMO radar is established. Then, the target localization is performed using a forward-backward spatial smoothing (SS) based linear prediction-orthogonal propagator method (LP-OPM) to obtain information about the radar target. Next, the YOLOv7 and the visual ranging algorithms are used to identify and localize the target to get information about the camera target. In addition, a Kalman-weighted fusion algorithm is used to fuse the data from the two sensor targets for output. Finally, the performance of the method is proved to be superior by the experimental results.

Keywords

• FMCW MIMO radar
• kalman-weighted fusion algorithm
• SS-LP-OPM algorithm
• YOLOv7 algorithm