Dynamic depth estimation of weakly textured objects based on light field speckle projection and adaptive step length of optical flow method

Repetitive textures in the depth estimation of weakly textured surfaces lead to numerous errors. The existing technology does not provide a complete method for light-field depth estimation in weak-texture scenes. Therefore we developed an algorithm that estimates the cost volume using speckle stereo correspondences. The subpixel-matching result is obtained using the adaptive label step length displacement theorem. The proposed method can adaptively select an appropriate step length for the sub-pixel phase-shift theorem without depending on the dataset category. The build cost volume can be adapted for weakly textured scenes. A common dataset was used to evaluate the effectiveness of the optical flow method for the adaptive selection of step length. Finally, high-precision step gauge blocks with a step difference of 10 mm were used to verify the depth-direction errors, which were less than 0.35 mm.