Study on identification of altered rock in hyperspectral imagery using spectrum of field object

•We identified the dominant object in the pixel using the spectrum of field object.•Spectral matching considered both the overall and partial shapes of the spectrum.•Altered rock was identified in the hyperspectral image of a forest-covered area.•Recognition result was shown by figures of altered rock and correlation coefficient.

In this article, the spectrum of a field object was used to match the pixel spectrum in the hyperspectral imagery, and determine whether there existed a dominant object in the pixel and to what extent the dominant object was present. Based on these recognition results, altered rock in the hyperspectral imagery was identified. The spectrum of a field object was obtained from field surveys in the study area, and in the spectrum, there was only one main object that could be effectively identified; this main object was named as the dominant object in this paper. In this study, the overall shape of the spectrum was matched firstly, and then the correlation coefficients between the pixel spectrum and the spectrum of field object in the corresponding wavelength ranges of the absorption-bands were calculated to determine the matching effect of these spectra as well as the possibility that the dominant object existed in the pixel. The recognition results were expressed by two images: the first image was a figure that illustrated the dominant objects in the pixels; the second image was a figure showing the correlation coefficients. The study area was a primitive forest covered region located in the Pulang porphyry copper mining area, Zhongdian County of Yunan province, China. The figures illustrating the distribution of the altered rocks and the correlation coefficient were obtained from this study. The on-site verification showed that using the spectra of field objects, it was possible to identify the dominant objects in the pixels, which favored the identification of altered rocks in the image. Moreover, compared to considering only the overall shape, the recognition results obtained by considering the overall and partial shape of the spectra had higher reliability.