This paper describes an algorithm of a change detection on the land using the different features data.
The conventional change detection method is roughly divided into two classes as follows:
(1) multispectral image comparison
(2) post-classification comparison
The change detection has to use two temporal observed images or classified data such as Thematic Mapper images and land use maps.
In general, the past data are the coded data such as land use maps or classified data, and the recent gathered data are the multispectral data. But there is the different sensor's specification data even if both of them are multispectral. Then an unadaption of specifications occurs in detecting changes between the different features data.
These data specifications are not satisfied in conventional methods. It may be impossible to detect the land cover changes.
In this paper is proposed a change detection method which is an effective comparison of thedifferent kind of specific characteristics data by incorporating the pseudo-spectral data. The pseudo-spectral data are given by adjusting the spectral characteristics between the compared data.
The proposed change detection processing is constructed with pseudo-spectral calculation which transforms the coded data into the pseudo-spectral data and the features extraction in the spectral feature space.
We show that the change detection accuracy of about 56% between categories of similar spectral features and about 90% between categories of dissimilar spectral features are obtained by the experiments using LANDSAT TM image.