THE EFFECT OF THE THERMAL INFRARED DATA ON PRINCIPAL COMPONENT ANALYSIS OF MULTISPECTRAL REMOTELY-SENSED DATA

The physical interpretation of principal component analysis of multisp ectral imagery has been established for reflective multi-band remotely -sensed data. Attempts to include the thermal data have not yielded an improvement in classification results despite the fact that the therm al data carries additional unique discriminative information as it dep ends on the bulk properties of the ground composites as well. This iss ue is studied by using high resolution groundbased images and previous ly published data. The inclusion of the thermal data disturbs the orig inal spectral composition of the eigenvectors set only if two conditio ns are met, which is frequently the case. The first is the existence o f at least one eigenvalue whose magnitude is approximately unity. If s uch eigenvalue is indeed present, then the corresponding eigenimage mu st have a substantial correlation with the thermal image, and this for ms the second condition. The second ranked principal component which i s associated with the typical reflectance of vegetation frequently ful fills these conditions and hence is affected by the inclusion of the t hermal data. Some methods to enhance the utility of joint analysis of reflective and thermal remotely sensed data are presented and discusse d.