Kazuma Natsume

Road-Shape Recognition Using On-Vehicle Millimeter-wave Radar

In this paper, we present a method for estimating road shape using on-vehicle millimeter-wave radar. The radar detects not only the backscatter from vehicles but also from stationary objects such as crash barriers and sound barrier walls that are generally installed along roads. This indicates that we can estimate road shape by collecting information on the distribution of these objects. We adopt the multiple signal classification (MUSIC) algorithm for increasing the azimuth resolution. Moreover, to apply the MUSIC algorithm into real-time radar systems, the computational cost is reduced by introducing the technique of bi-iteration singular value decomposition (Bi-SVD). The estimated road shapes exhibit good agreement with the actual shapes. Therefore, this technique is expected to improve the reliability of driving support systems.