Object modelling based on laser metrology and neural networks

Abstract

An automatic technique for shape modelling is presented. The approach of this technique is the object representation by means of a mathematical model. To carry it out, a neural network is implemented to perform the object modelling. The model provided by the neural network is performed based on the surface data. To detect the surface data, the object is scanned by a light line. Thus, the laser metrology provides the surface data to construct the model by means of the network. This process involves image processing of a laser line pattern. The approach of the neural networks is to perform the object modelling without measurements on the optical setup. Thus, the setup performance and accuracy are improved. It is because the errors of the measurement are not added to the computational model. To describe the accuracy a root mean square of error is calculated using data provided by the network and data given by a contact method. This technique is tested with real objects and its experimental results are presented.