Title :
Application of feedforward neural networks to object recognition for image analysis
Author :
Sigillito, Vincent G. ; Sadowsky, J. ; Bankman, Isaac N.
Author_Institution :
Appl. Phys. Lab., Johns Hopkins Univ., Laurel, MD
Abstract :
Summary form only given, as follows. The automated analysis of X-ray images of mechanical fuses for nondestructive evaluation (NDE) requires that objects within the images be detected and recognized and that the orientations and relative positions of these objects be computed. Neural-network-based adaptive spatial filters for performing these functions are advantageous for many reasons, including robustness, flexibility, and adaptability to changes in fuse design. A research project has been started to investigate the use of feedforward neural networks to improve the performance of an NDE image analysis system. In particular, the objective was to develop a system which learns the rules for image understanding and could be applied to new or modified fuse designs with a minimum of software modification
Keywords :
adaptive filters; computer vision; computerised pattern recognition; electric fuses; neural nets; nondestructive testing; radiography; spatial filters; NDE image analysis; X-ray images; adaptability; adaptive spatial filters; feedforward neural networks; flexibility; image analysis; image understanding; mechanical fuses; modified fuse designs; nondestructive evaluation; object recognition; orientations; relative positions; robustness; Feedforward neural networks; Fuses; Image analysis; Image recognition; Neural networks; Object detection; Object recognition; X-ray detection; X-ray detectors; X-ray imaging;
Conference_Titel :
Neural Networks, 1991., IJCNN-91-Seattle International Joint Conference on
Conference_Location :
Seattle, WA
Print_ISBN :
0-7803-0164-1
DOI :
10.1109/IJCNN.1991.155524