A framework for 3D tracking of frontal dynamic objects in autonomous cars

Both recognition and 3D tracking of frontal dynamic objects are crucial problems in an autonomous vehicle, while depth estimation as essential issue becomes a challenging problem using monocular camera. Since both camera moving, the can be formed structure from motion (SFM) problem. In this paper, to elicit features image, YOLOv3 approach is utilized beside OpenCV tracker. Subsequently, obtain lateral longitudinal distances, nonlinear SFM model considered alongside state-dependent Riccati equation (SDRE) filter newly developed observation model. Additionally, switching method form error covariance proposed enhance robust performance SDRE filter. The stability analysis presented conducted on class discrete systems. Furthermore, ultimate bound caused by uncertainties analytically obtained investigate significance. Simulations reported validate switched Finally, real-time experiments performed through multi-thread framework implemented Jetson TX2 board, radar data used for evaluation.