DISTRIBUTED IR SENSOR ARRAY FOR OBJECT CLASSIFICATION

Boris Ivanov, Vladislav Pavlov, Heinrich Ruser

Abstract

This paper addresses the problem of recognition and identification of objects with an IR diode array, working on a reflection light scanner principle. Essentially two arrays comprising of 3 emitter-receiver pairs are mounted on two sides of the area of inspection, enabling the estimation of the size of the object in different dimensions and reducing the requirements with regard to the detection range. The sensors are driven successively in time, hence no signal overlapping and cross-talk occur. For the recognition, a neural network approach based on the Backpropagation algorithm has chosen. The array data are preprocessed via a Principal Components approach. As a result various objects can be recognised and classified easily and are well separable from other echoes. This work is preliminary for a practical system determining the number of people and identifying people getting into or out of a room and other applications supporting important home appliances like occupation-driven HVAC control or determining behaviour patterns.

Keywords

IR multi-sensor array, object classification, neural network approach

Download

IMEKO-TC7-2004-110.pdf

DOI

-

IMEKO TC

TC7 - Measurement Science

Back to the proceedings