Research on multi baseline large depth 3D imaging

The acquisition of 3D information about targets of interest in real scenes and 3D scene restoration is receiving more and more attention in the field of machine vision. Binocular stereo vision techniques are already relatively mature for this purpose. Traditional fixed-baseline binocular stereo imaging techniques are more accurate in 3D imaging of targets at close range. However, in the face of large depth scenes, the 3D imaging effect in terms of imaging accuracy and imaging distance can no longer meet the demand. In this paper, we use the multi-level and multi-view characteristics of multi-baseline 3D imaging to obtain 3D information of targets of interest at different depths in the range of 5m-15m depth of field. The error of different depth-of-field targets in large depth scenes is maintained at a controllable level, thus reducing the overall error of 3D imaging of targets in large depth scenes.