To standardize athletes'' movement postures， a method for correcting athletes'' postures based on adaptive modeling of an improved visual background extractor （ViBe） is proposed. First， a ViBe feature descriptor for non-standard postures is defined， and adaptive modeling is implemented for the improved ViBe feature descriptor to mark the non-standard postures of athletes， thus achieving non-standard posture detection based on the improved ViBe adaptive modeling. Then， a non-independent coordinate system is established， and the movement posture within the dynamic closed loop is corrected through the coupling of posture deviations and completing the design of the athlete posture correction method. Finally， MATLAB-Adams software is used for simulation experiments， taking robot movement as the research object， and selecting jumping movement as the target posture for experimentation. The results show that under the action of the above correction method， the hip joint of the tested robot no longer rotates， and the rotation angles of the main leg joint， the secondary leg joint， and the foot joint can also be effectively controlled， indicating that this method can effectively correct athletes'' postures.