Determining the quality of surfaces by utilizing scattered laser light that is examined with a neural network is discussed. Features are extracted from the scattered angular spectrum and then used as inputs to a hierarchical neural net. The net is trained by a selected set of machined surfaces whose quality has been independently established. These samples are repeatedly presented to the sensors, and the network will, each time, make a decision about the surface roughness which is then compared to the correct answer. The error is used to modify the connection weights. Following this training period, the net is able to identify the quality of surfaces presented to it, even in the presence of noise caused by poor illumination. Experimental results from a set of lapped surfaces are discussed with regard to the fusion of different features in order to obtain an adequate measure of surface roughness using the neural network. It is shown that the system is able to interpolate between the surfaces of different preparation and estimate the roughness parameter in the micron age.<<ETX>>
[1]
Stephen Grossberg,et al.
Art 2: Self-Organization Of Stable Category Recognition Codes For Analog Input Patterns
,
1988,
Other Conferences.
[2]
Richard P. Lippmann,et al.
An introduction to computing with neural nets
,
1987
.
[3]
Theodore V. Vorburger,et al.
Surface Roughness Metrology By Angular Distributions Of Scattered Light
,
1985,
Photonics West - Lasers and Applications in Science and Engineering.
[4]
Yoh-Han Pao,et al.
Adaptive pattern recognition and neural networks
,
1989
.
[5]
Theodore V. Vorburger,et al.
Optical Roughness Measurements Of Industrial Surfaces
,
1986,
Other Conferences.
[6]
Oskar Gerstorfer,et al.
Optical Roughness Measuring Instrument For Fine-Machined Surfaces
,
1985
.
[7]
A. Wirgin,et al.
Scattering from sinusoidal gratings: an evaluation of the Kirchhoff approximation
,
1983
.