Multidisciplinary analysis for feature information extraction of Remote Sensing

Multidisciplinary analysis for feature information extraction of Remote Sensing

Chi Guobin, Li Yan
Xinjiang Institute of Geogrpahy

Din Xuan
Guiyang Institute of Geochemistry

Tond Qingxi
Aero borne Remote Sensing Center, Chinese Academy of Sciences.

Introduction
Being one of the important methods of Remote Sensing, the image processing is becoming widely used in many fields. With various functions possessed or new developed in image processing system, the procedure of enhancing certain information interested (object information), enabling them to be easier identified from background information is used for extraction of feature information. However, there are still questions whether the extracted information coincides with the expected information how about its reliability, what mechanism is represented by the feature information, and whether the inherent laws of object information are reveled, these are directly related to the results of image processing and its practical value.

Consequently, in order to obtain good result with practical value, it is necessary not only to study the fundamental features of the object information in detail, work only a feasible plan for conduct the verification for the extracted feature information. We call this whole set of analyzing method as multidisciplinary analysis", its major steps are shown as follows :

Fig. 1 Flow-chart of multidisciplinary analysis
Regional geology and spectral features of major geobodies
Tiemuerte region is located in the Kelan synclinorium of Altai fold system according to the geotectonic units. In mining district, exposed strata are mainly composed of Kangbutiebao group of lower Devonian series, and its is classified into for lithological units which stenches from northeast to southeast. Polymetalic mineralizetion is generated in the third lithological unit which mainly consist of tuff, marble, quarts biotite schist, chlorite quartz schist, skarn, calcsiltstone etc the research work proved that polymetica mineralization have a direct bearing on skarn and impure manganese carbonate, the expouring gossan may reach a width of 25-50m, even hundreds meters on the surface.

Since 1985, for application of aeroborne FIMS technique to the geological exploration a group of scientists of Chinese Academy of Sciences began their research works in Tiemeurte, Altai, Xinjiang. Simultaneously, with airborne data acquisition by 6 channels infrared multispectral scanner between 2.0..2.5 micrometer spectral region the field spectral reflectance measurements were made in same spectral region at same area. Spectral bands used for FIMS are as follows :

2.035mm
2.087mm
2.14mm
2.200mm
2.280mm
2.380mm

The scanning range of down trip including II, III and IV lithological units, and samples of main rocks in these three lithological units were measured and analyzed in laboratory. Almost all basic data were collected to support the analysis of FIMS data processing and interpretation of major geobodies in Teimuerte region.

Comparison have been made between the data measured in the laboratory (See Fig. 2) and on the feild (see fig. 3) with the samples of mica-quarts schist, gossan and quartz sodaclase porphyry. It belongs to different lithological units in turn of the second, the third and the fourth one.

Fig. 2 Laboratory measured continous spectrum of typical rocks from lithological units to in Teimuerte

Fig. 3 Laboratory measured 6 channels spectrum of typical rocks from lithological units to in Teimuerte
In the whole region of six bands the geobodies's reflectance in lithological units II and IV are higher than in unit III. Speaking in general terms, bands 2 and 3 mainly reflect the characteristics of high reflectance of geobodies ranging from lithological units II to IV in Teimuerte area. And bands 4 and 6 reflect theirs low reflectance characteristics in absorption.

Image processing and extraction of mineralized feature information
The principal component analysis in image processing method by mapping the, digital image information in band space to the selected principal component space for identifying the features information. It was seen from the mechanism study of principal components analysis that principal components and constituted by linear combination of every band, and each principal component including the information of every bands. The enhancing and merging information exists on each principal component, but the amplitudes of enhancement and merge are different. Each principal component contains the different information types of enhancement and merge, the principal component analysis would show the effect of comprehensive enhancement after recombining the intrinsic information. Therefore, the principal component image of color composition would improve in greatly degree of the information identification or the area interpretability.

There are two steps with the extraction of FIMS image of down strip in teimuerte. The purpose of this interference is to discriminate three lithological units in down strip and extracted mineralized information in the lithological unit III. Though the two steps used the same bands (bands 2+3, 4, 6) and method (principal component analysis) the division of three lithological units mapped in a distribution direction and position of band combination which reflect several main geobodies in unit III, the extraction of mineralized feature information was based on the steps above. its mapping transformation depended on distribution on Gossan in band combination. One of the research is shown in place I. The right half is a color composition image of principal components of study are in which three different color belts in the black background are clearly showed. The left half is a processed image of extraction of mineralized feature information.

In the left image, the whole background is homogenous blue color tone, white and yellow color lumps, yellow-green color plaques, and pinkish and light pinkish color spots are distinguished on image. It may be seen by comparing with right half of the image that the whole image is in blue color except the white plaque with some pinkish color in lithological unit II, it is mainly in white and yellow color lumps and a little yellow green color plaques in unit III, it is mainly in white and yellow color lumps and a little yellow green color plaques in unit III, it is blue also in unit IV except little yellow, and yellow, and yellow green plaques. The analysis shows that the merge of large amount overlaid information are in blue color, and the yellow section in form of lumps is easily distinguished from surrounding information which is in consistent with it. The position of these yellow lumps on middle up per of unit III and contact zone between units III and IV. This result is exactly the same as mineral zone in the layer of the region.

Verification of Mineralized Feature information
In the research work, the procedure from check to establishment is called "verification analysis" we employed exactly the same method used in image processing system with the data related to unit III, all the work of calculating was performed on microcomputer.

First, 53 pixels in a small rectangular area were selected from image for principal component analysis. For the direct perceiving, their distribution values in principal component space were drawn on the plane which was constituted by the first and the second principal components. According to the principal component values of these pixels which were obtained from image processing, they were calculated by CIE chromatic map. In all these 53 pixels, most of them were in yellow, light green - yellow, yellow green and light yellow-green. As the color tone of each pixel, a boundary different color large section with two parallel lines. The upper of the double line is in yellow color. The transition belt is in yellow color. The transition belt is in yellow green between two lines. Below the single line is in green color tone mainly. The boundaries of each color tone are obviously on the plane of principal components, without blending pixels of different color tones, except No. 4 and No. 37 pixels in yellow green color tone, and both of which fall into the extent of green yellow.

For image processing it is, a key problem to study the practical meaning of yellow abnormalous or to analyze by which rocks and mineral spectral features the abnormal color tone is caused. The results of which was calculated from 30 spectral samples of six main rocks and minerals in unit III was overlaid. It may be seen that the values of principal component for most of the samples, such as belt sharped gossan skarn and garnet chlorite wuartsz schist, are distributed within the extent of yellow color tone. The samples numbered ,9 and 8 are distributed in the extent of light green- yellow where is near the yellow extent. Only one of sample of garnet chlorite quarts fallen into the yellow green extent.

The sample of chlorite quarts schist almost distributed in yellow-green extent, and tuff and impure crystalline limestone are fall into the extent of green color tone. In a word, the overlaying analysis of pixel color tone and spectral feature corresponded show that the mineralization is represented by the yellow abnormal color tone.

For confirming the corresponding relationships of abnormal color tone and mineralization, the principal component value of chemical component analysis data calculated from 13 rocks samples which collected in transition section between units III and IV was overlaid on the extent of pixel's color tone (fig. 4) it can be easily seen that belt sharped iron ore, gossan and skarn are distributed in yellow area, metaquartzite and quartzite are distributed in yellow green extent, tuff and impure crystallines limestone are distributed mainly in light green yellow extent, since they were collected from transition section of units III and IV with lower degree of mineralization. Number 4 quarts biotite schist is in yellow extent as it has higher content of ferrous oxide.

Fig. 4 The overlapping diagram of chemical composition analysis and pixels's extent of rocks and ore
Making comprehensive survey of the result by principal component analysis respectively with the data of image processing, spectral data and chemical composition data coincide much better with each other, and the changing tendency is in full agreement with color tone, content of chemical composition and mineralized degree. In all respects, the yellow abnormal color tone actually show the mineralized feature information. Mineral belt itself and intense altered rocks such as skarn and garnet quarts schist are to be shown in the yellow color tone on image, less altered rocks, such as chlorite quarts schist, shows the transitive color the between yellow and green. Rocks without mineralization, such as tuff, impure crystalline limestsone, shows the green color tone mainly.

Mechanism Explanation : Geological Significance of Mineralized feature information
It is discovered by the research work there is a closer corresponding relationship between principal component value of chemical components of rocks and minerals and its content in percentage. In order of percentage content from large to small, the content of silica with the first principal component value are listed in table. 1

Table 1 corresponding table between chemical composition content of rocks and ore with principal component value.

From table 1, the content exactly correspondent with the highest and lowest value of principal component values. F1 mainly reflects the change of silica and ferrie _______________ reflects mainly changes of ferrous oxide consent, in Fig. 4 the samples of rocks and mineral distribute on the yellow area, their chemical composition feature is in lower consent of silica (as the order of the seventh and tenth to the thirtieth) but highest (as the order of 1st , 2nd, 3rd, 5th and 7th) ferri-ferrous (ferricoxide plus ferrous oxide). As mentioned above, the color tone changes form yellow to green from upper left to lower right on the plane of F1 and F2, the yellow are s the area of lower value F1 but higher value F2, the yellow area is the area of lower value F1 but higher value F2, th - is means that the mechanism of mineralized feature information shown in yellow abnormal color is the mineralized feature of lower silica but much higher content of ferrous oxide. As for polymetalic mineralization belt to Tiemuerte, the yellow abnormal color tone in the form of lump shown the distribution of gossan, magnetic and skarn which are exposuring on the earth's surface.

Conclusions
From the above analysis we can conclude as follows :

1. There are four steps of comprehensive analysis for extracting prehensive analysis of Remote Sensing, which are independent of each other and also related to each other. The steps include : The analysis of essential feature information, image processing, verification analysis and mechanism explanation. Verification analysis is an important steps to discuss whether the feature information could be established, and to analyze the inherent significant.

2. The method of principal component analysis is not only efficient in extracting feature information but also in overlapping multiple information (including non-Remote Sensing information). It is the most important method in multidiscplinary anlaysis

3. The yellow abnormal color tone shown in the mineralized feature information of narrow-band IR flight strip o Tiemuerte in image processing result. It reflected on mineralized feature with lower content of silica and higher ferrifcferrous content in polymetalic mineralization belt of the region. Yellow abnormal color tone in form of lumps shown the distribution of gasson, magnetite and skarn on exposing surface. According to field examination, the distribution section of yellow color tone is in agreement with the mineralized area on the spot surface.

References

S. Wenderoth, E. Yost et. al 1975 MULTISPECTRAL PHOTOGRAPHY FOR EAR THE RESOURCES. Remote Sensing information center greenvale, New York.
Hunt G.R, J.W. Salisbury and G. J. Lenhoff, 1971, Visible and Near-infrared spectra of minerals and Rocks: Oxides and Hydroxides; modern geology, vol. 2, PP 1950205.