Depth Estimation With Structured Light Field Sub–Aperture Fringe Images

The acquisition of scene depth data is a classic problem in the field of computer vision.Digital projection fringe measurement technology has become a research and application hotspot in the field of depth estimation due to its remarkable advantages such as non-contact,high precision and fast calculation.Conventional digital projection stripe depth acquisition systems often use CCD/CMOS cameras as acquisition devices,but traditional systems have limitations in depth data acquisition of objects with high dynamic surface or large depth of field.But the combination of light field camera and digital projection stripe also brings new challenges.We explored the following two challenges in this paper.First,since the light field sensor simultaneously records the intensity and angle information of the light,the spatial resolution of the sub-aperture image is low,and the spatial resolution of the depth data calculated by the sub-aperture stripe image needs to be improved;secondly,there is sub-pixel offset between sub-aperture images.The accuracy of existing matching-based offset estimation methods is not high.Inaccurate offset data will have a significant impact on the fusion of sub-aperture depth data.We proposed a sub-pixel offset calculation method based on the absolute phase of the projected fringes is proposed.Then the POCS(projection on convex sets)algorithm is used to fuse the multiple sub-aperture depth maps.A depth estimation result with higher precision is obtained.In order to verify the effectiveness of the proposed method,different objects were selected for experiments: First,the standard gauge block was measured.The minimum identifiable depth was 0.2mm when the camera lens was 36 cm away from the reference plane.2)Secondly,the multiple sub-aperture depth data fusion experiment was carried out on the computer keyboard image.The experimental results prove the effectiveness of the method.3)Finally,experiments are performed on two figures with complex surface topography,and the high precision depth is obtained.The experiment result demonstrates the depth data acquisition capabilities of the proposed method.