Fig. 3: Analysis of 3D imaging in scenes with reflective surfaces.

(a) and (b) are scheme and picture of the indoor glass wall scene. The glass wall is about 1.2 m in front of the sensor, with a reflecting image of two persons and the sensor. Note that two persons are on the same side as the sensor. c Depth image from the stereo vision camera (Intel RealSense D455). d Depth image from the PSL 3D sensor. The glass part is denoted by the dashed line. (e) and (g) are testing scenes of the deep learning method. Their corresponding prediction results are shown in (f) and (h). The jet color bar applies to (c) and (d), the gray color bar applies to (f) and (h).