Computer vision of the foot sole based on laser metrology and algorithms of artificial intelligence

Abstract

An automatic technique for the 3-D vision of the foot sole is presented. This technique is performed by means of laser metrology and approximation networks. To retrieve the topography, the foot sole is scanned by a laser line through a glass window. The contouring of the foot sole is based on the behavior of the laser line. This 3-D modeling is performed by an approximation network. The structure of this network is built based on the line shift that is generated due to surface variation and the camera position. Also, the intrinsic and extrinsic parameters of the vision system are computed based on the network. In this manner, online setup modifications can be performed. Thus, the external measurements are not passed to the vision system. In this manner, the accuracy and the performance are improved because physical measurements are avoided. The approach of this vision system is to fit the shoe sole mold to the foot sole via contour curves. The results are evaluated by means of a root mean square of error using references from a contact method. Thus, a contribution in computer vision is achieved for profitable shoe design. The processing time is also described.