A neural network based histogramic procedure for fast image segmentation

Author

Kothari, Ravi ; Klinkhachorn, Powsiri ; Huber, Henry A.

Author_Institution

West Virginia Univ., Morgantown, WV, USA

fYear

1991

fDate

10-12 Mar 1991

Firstpage

203

Lastpage

206

Abstract

The determination of the dimension of a lumber board, the location and extent of surface defects on it, are essential in the construction of a visual inspection station for the lumber industry. The paper presents a neural network based histogramic procedure that performs on the image of a board and can be used to determine the board dimension, the location and extent of surface defects on it, in near real time. The method is based on segmentation of the image based on multiple threshold information derived from a multi-layered neural network. Such a scheme can be applied in general to image analysis and the implementation shows fast processing requiring very little control over the environment. The construction of the network and its training are also discussed

Keywords

computerised pattern recognition; computerised picture processing; inspection; neural nets; word processing; computerised pattern recognition; computerised picture processing; fast image segmentation; image analysis; lumber board; multi-layered neural network; neural network based histogramic procedure; surface defects; training; visual inspection station; word processing; Cameras; Colored noise; Image segmentation; Laser beam cutting; Machine vision; Neural networks; Noise generators; Pixel; Wood industry; Working environment noise;

fLanguage

English

Publisher

ieee

Conference_Titel

System Theory, 1991. Proceedings., Twenty-Third Southeastern Symposium on

Conference_Location

Columbia, SC

ISSN

0094-2898

Print_ISBN

0-8186-2190-7

Type

conf

DOI

10.1109/SSST.1991.138548

Filename

138548