RECONHECIMENTO DE PADRÕES DAS IMAGENS QUIMICAS GERADAS POR SENSOR DE GÁS TIPO CAPACITOR MOS

Abstract

This paper reports the result of pattern recognition and classification of gas response of a MOS chemical image sensor. The chemical images were obtained by scanning pulsed light on the sensitive surface of the sensor. With this sensor were analyzed responses of H2 and NH3 gases at different concentrations. To perform pattern recognition analysis of images, two different chemical methods have been proposed, co-occurrence matrix and histogram of intensities of gray levels methods, from which they were extracted characteristic parameters of images. The final classification process was performed by principal component analysis (PCA) of the characteristic parameters. The proposed methods were satisfactory and robust to the concentration of the gases.

Index Terms — Chemical Images, MOS Sensor, Pattern Recognition, PCA.