Estimation of NLOS Obstacle Position Using Reflected Image on Transparent Surface

Abstract

In this paper, a method for estimating the position of obstacles existing outside the line of sight of the vehicle by using the reflected image on a transparent surface is proposed. The proposed method takes two pair of pictures that obstacles reflected on a transparent surface with a stereo vision camera: one is the pair of image taken directly and another is the image taken through a polarizing filter. A reflected image is extracted by suppressing the background image transmitted through the transparent panel from the difference between these images. The disparity information of these left and right reflected images is taken, and the distance to the obstacle is estimated. Distance estimation experiment was conducted with three patterns of the incident angle of the reflected image. The experimental results indicated that it was possible to estimate the distance without considering the influence of the transmitted background light if it is the reflecting surface existing in the adjacent lane. On the other hand, it shows that the proposed method is effective when using the reflection on the transparent surface about 5 m away from the road.