This invention relates to the field of 3D
human body reconstruction technology and discloses a method for adjusting the abdominal size of a 3D
human body model based on
image processing and SMPL (Small Scale
Programming). It uses the edge difference between the 2D side view contour and the model's side view contour as a driving quantity, converting pixel-domain differences into the maximum
magnification distance at the real scale. Furthermore, it utilizes the
attenuation coefficient of vertices within the abdominal region to achieve continuous outward extrapolation from the center to the periphery and from key abdominal positions to the upper and lower boundaries. This allows for zero-sample adjustment of abdominal size without altering the SMPL reconstruction process or retraining the model, improving the modeling accuracy for pregnant women and those with
abdominal obesity, and reducing the cost of reconstructing obese individuals. Therefore, this invention achieves targeted adjustment of abdominal size without retraining the model or collecting large-scale training data, improving the 3D reconstruction accuracy of the
abdomen in obese individuals while also maintaining modeling accuracy for individuals with normal body types.