Three-dimensional profile measurement by using an artificial neural network

This paper introduces 3D profiles noncontact measurement method by using a kind of artificial neural network- backpropagation network. It makes a detail discussion on optical system, data acquisition, and how to train the BP network. An object is projected by a laser stripe light, and the lien image of the object is captured by a CCD camera. The input data of BP network is given from a series of image on CCD by certain image processing technology. Once a mapping relationship of coordinates between CCD camera image plane and related position in object's space is set up by sing BP network, the image plane coordinates of any objects are used as the input data of BP network it will immediately obtain the corresponding coordinates in space according to the output. Thus the profile can be easily established by these coordinates. This method reduces the requirement for system accuracy, and no need to test and a light the whole system in advance accurately.© (1998) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.