[1N3-GS-10] Vision, speech: GAN

[1N3-GS-10-03] Dense 3-D Visualization of Underground by Interpolating Penetrating Radar images with GAN

Yasushi Kanazawa¹, 〇Kaito Tokushige¹, Jun Sonoda², Yoshihito Yamamoto³, Kazutaka Mitsutani³ (1. Toyohashi University of Technology, 2. National Institute of Technology, Sendai College, 3. Nagoya University)

Keywords:Image Interpolation, Penetrating Radar Image, GAN

We propse a method for dense 3-D visualization of underground information by interplating ground penetrating radar images. A ground penetrating radar can only measure the condition of the underground directly below a survey line. So, for dense measurements, it is necessary to narrow the interval between survey lines. In this paper, we generate intermediate images between survey lines by interpolation with GAN. By using the generated images, we can reduce the cost of the measuring and visualize dense 3-D information of underground. We have shown that our method can generate more accurate interpolate images than linear interpolation.