A technique for 3D shape detection based on light line image processing and neural networks is presented. The technique consists in the projection of a laser light stripe over the object. The light line then is distorted by changes on the object surface. The relief of the object is obtained by measuring the displacement of the light line. A neural network is implemented with data from line displacements corresponding to known object heights. The neural network models the behavior of the displacement of the laser line over the objects. In this way, the parameters of the experimental setup are not used and the results are improved. The performance of the technique is evaluated with the rms error, which is calculated by using data from a Coordinate Measuring Machine and simulated data.
3D Object Reconstruction Using Structured Light and Neural Networks
2010-09-01
1129267 byte
Conference paper
Electronic Resource
English
Efficient 3D data fusion for object reconstruction using neural networks [3647-11]
British Library Conference Proceedings | 1999
|