Estimation of the orientation of vehicles in high-resolution radar images

Abstract

The availability of high-resolution image radars allows estimating the orientation of vehicles from a single measurement without temporal filtering. This gives the opportunity to react even faster to certain critical traffic scenes. This paper presents an approach for estimating the orientation of a vehicle. The orientated bounding box algorithm known from literature is adapted to this end and a quality function is introduced to choose the optimal bounding box. In addition, a brute-force approach for determining the best possible outcome is presented.

Radar Sensors for Autonomous Driving: Modulation Schemes and Interference Mitigation

“…The dimension, orientation, and motion state of target vehicles can be used by an autonomous vehicle to plan its driving manoeuvres. High-resolution radar sensors can extract the dimension and orientation of target vehicles [15,16], cf. Fig.…”

“…The x-y representation of the target vehicle with estimated orientation ( ) as described in[15,16]. (c) Camera image of the measured scenario.…”

Template matching for radar-based orientation and position estimation in automotive scenarios

“…Currently, most of the research regarding position and orientation estimation based on radar is limited to simplistic approaches like that of an Oriented Bounding Box (OBB) [1].…”

Adaptive clustering for contour estimation of vehicles for high-resolution radar

“…Up to this time, only a few investigations on the topic of radar based contour estimation exist. In [2] an algorithm chain for obtaining an oriented bounding box based upon a density based clustering algorithm [3] is presented.…”

“…The algorithm chain in this paper is based upon the approach in [2], but uses an adaptive clustering algorithm presented in [4]. This KNN-DBSCAN algorithm is better equipped for self-configuration and adaptation to different radar data characteristics with respect to reflection point distribution and density.…”