Depth information processing method based on visual reconstruction

The invention discloses a depth information processing method based on visual reconstruction, and the method comprises the steps: 1, enabling a robot to adjust the luminosity consistency cost value of a pixel point through setting the depth information adjustment amount, and recognizing the type of a texture region where the pixel point is located according to the adjusted luminosity consistency cost value; step 2, when the robot recognizes the pixel points in the first texture area, the robot processes the depth information of the pixel points in the first texture area by using a luminosity consistency constraint criterion; when the robot identifies pixel points in the second texture area, the robot processes depth information of the pixel points in the second texture area by using cost information of luminosity consistency and a geometric projection consistency constraint criterion; and step 3, the robot performs inter-frame normal vector consistency constraint by using the depth information newly processed