Representation of Shape Based on Laser Stripe Projection and Modelling with Bezier Approximation Networks

摘要

An automatic technique for the representation of the object shape is presented. The main aim of this technique is the extraction of the topography and the shape of articles by networks. Based on laser stripe projection, the topography is obtained automatically using Bezier approximation networks. Using the topographic dimension, a two-dimensional Bezier approximation network is able to calculate the shape of an object. The topography is detected by scanning the object by the laser stripe. The scanning produces a set of images of the laser stripe, which are processed by Bezier approximation networks to determine the object dimensions. The Bezier approximation network calculates the dimensions according to the behaviour of the laser stripe. Based on these dimensional data, the Bezier network produces an object model. This model is a replica of the object shape. The idea behind the Bezier approximation networks is to obtain the shape of an object without any measurements being carried out on the optical setup. Thus, the performance of the experimental setup and the accuracy of the results are improved. This is due to the fact that errors of measurement are not passed on to the computational model. The accuracy is assessed by taking the root mean square of the error and data produced by Bezier approximation networks and the data obtained by a contact method. This technique is tested with real objects and the experimental results are presented and assessed.