Accession Number : ADP009013

Title :   Neural Net for Trajectories Recognition in a Flow,

Corporate Author : UNIVERSITA DEGLI STUDI LA SAPIENZA ROME (ITALY)

Personal Author(s) : Cenedese, A. ; Romano, G. P. ; Paglialunga, A. ; Terlizzi, M.

Report Date : 23 JUL 1992

Pagination or Media Count : 5

Abstract : In order to recognize the trajectories from a sequence of single exposure images a new method, based on the simulation of a neural network, is proposed. The net is identifies by the topology, the nodes characteristics and training rule which allows a change in the weight of each connection. The method is utilized to analyse the trajectories described by tracer particles flowing inside a porous medium simulated by means of fractured Pyrex glass bars. The optical access is realized using a liquids mixture with the same refractive index of the solid matrix (matching index technique). (Author)

Descriptors :   *FLOW FIELDS, *IMAGE PROCESSING, *NEURAL NETS, TOPOLOGY, WEIGHTING FUNCTIONS, DIFFUSION, POROUS MATERIALS, OPTICAL TRACKING, TRACER STUDIES, FINITE DIFFERENCE THEORY, ITALY.

Subject Categories : Fluid Mechanics
      Computer Programming and Software
      Recording and Playback Devices
      Cybernetics

Distribution Statement : APPROVED FOR PUBLIC RELEASE