An automatic technique to improve the plantar surface model via Bezier networks is presented. This technique generates the foot sole model via Bezier networks based on surface points, which are retrieved via line projection. The surface model is defined by means of network weights and the control points, which are measured via line position by a Bezier network. Thus, the model represents the plantar surface with high accuracy. It is because the model passes through by all control points of the physical surface. Furthermore, the network reduces operations and memory size to calculate the surface. It is because the model is implemented with less mathematical terms than the traditional models. Also, the calibration of vision parameters is performed via laser line to avoid external calibration, which determines vision parameters outside of the vision system and increases the surface representation inaccuracy. Additionally, the shoe-last bottom is adjusted to plantar surface model. The viability of the proposed surface modeling is corroborated by an evaluation based on the speed, accuracy and memory size of the traditional surface models. Thus, the contribution of the proposed technique is elucidated.