|
Mining plants are one of the factors having major negative impact on the area where they are situated. In our study this is
the case of the mine production plant consisting of Elacite mine and Mirkovo floatation plant both located in central part
of Stara Planina Mountain. In this study an attempt is made to delineate the overburden dumps and open pit mines by
means of remotely sensed multispectral data with moderate spatial resolution (e.g. Landsat TM/ETM+ 30m) is a
challenging task. The major difficulties arise from: 1) large period using the dump (introducing the need for
multitemporal data); 2) the unknown proportions of vegetation, soil and embedding rock samples in the boundary areas
and their seasonal variations; 3) relatively restricted access to places of interest. A variety of methods have been
proposed to overcome the problems with pixels corresponding to two or more end-members, but a promising one is the
soft classification which assign single pixel to several land cover classes in proportion to the area of the pixel that each
class covers. In this scenario for every pixel of the data the correct proportion of the end-members should be found and
then co-registered with the corresponding original pixel. As a result this sub-pixel classification procedure generates a
number of fraction images equal to the number of land cover classes (end-members). The sub-pixel mapping algorithms
we have exploited so far have one property in common: accuracy assessment of sub-pixel mapping algorithms is not easy
because of missing high resolution ground truth data. One possible solution is to incorporate in the method adopted
additional ex-situ and in-situ measured data from field and laboratory spectrometers with bandwidth about 1 nm. This
study presents a successful implementation of soft classification method with additional, precise spectrometric data for
determination of dump areas of the copper plant and open ore mine. The results achieved are proving that the in-situ
gathered data provide coincidence of 93.5%. The main advantage of the presented technique is that mixed pixels are used
during the training phase. Compared to these other techniques, the present one is simple, cheap and objective oriented.
The results of this sub-pixel mapping implementation indicate that the technique can be useful to increase the resolution
while keeping the classification accuracy high.
|
