A Neural Network Based Shape-from-Shading Measuring Technique for Engineering Industries

Abstract

A new approach for measuring shape and surface of an object for engineering measurement is proposed. The proposed methodology is a neural network based Shape from Shading (SFS) technique. In this SFS approach, the physical parameters of the reflectivity under different lighting conditions are interpreted by the neural network weights. The proposed technique optimises a proper reflectance model by an effective neural learning algorithm. The depth of the object surface is recovered by using a simple SFS recursive algorithm. Experimental results are demonstrated that the proposed technique exhibits high efficiency and accuracy of measuring an object surface for the manufacturing industry.