GISdevelopment.net ---> AARS ---> ACRS 1990 ---> Poster Session Q

Integration of Remote Sensing and multi source geo data its application in structure analysis and mineral prospecting in Tongling, China

Yang Wenli, Wang Runsheng
Center for Remote Sensing on Geology
Ministry of geology & mineral resources PRC


Abstract
A variety of image processes and information extraction techniques are applied to the reregistered multi source image sets of Tangling area in East China including conventional image processing textural information extraction quantitative analysis of liniments image processing of potential fields data and symmetric processing of aromas and gravity data. Fracture structure controlling mineralization is analyzed according to the interpretation of integrated image sets. Variables derived from the image sets are used predict the potential mineral deposits in the study area. A characteristic space analysis method producing new variables moss effective on discriminating different mineralization conditions through orthogonal transformation are designed applied five promising areas are pointed out.

Introducation
There has been long history of copper ore exploration in Tongling area . Originally, mineral targeting was carried out mainly through seeking surface ore occurrences from 150 theoretical targeting predicting mineral deposits according to certain mineralization patterns has gradually become the dominating method in recent ten years more emphases have been placed on the use of multi source data to solve geological problem and to predict deposits but most work is still confined to the separate analysis of different data. Integrated image processing of remote sensing and multi source geodata is rarely reported in the area. In the study reported in this paper, multisource image sets are created and various information extraction techniques are adopted and developed. The effectiveness of these techniques in geo structure research and mineral prediction is discussed.

Geological out line
The main strata in the area are the marine faces or alter faces of marine and continent elastic rock form tics from Silurian to Devonian system and carbonate rock formation from carboniferous to Triassic period magnates and tectonics were very weak in the weak in the area from complicated folding and furcating of plate cover had formed movements copper deposits area closely tied with structure magnetic rock and strata magmatism and mineralization are controlled by fracture structure. Mineralization has close relation to intermediate to acid rocks . Carboniferous Permian and Triassic system are favorable strata for ore deposits.

Data processing
  1. Integrated image sets
    Integrated images sets consists of following data 1) Two MSS images ( Acquired on July 6 1978 and July 17 1981 ) one TM image ( acquired on Nov 15 1985 3) 1:50,000 aeromagnetic data; 4) 1:50,0000 soil geochemistry anomaly maps of Au Ag, Cu, Pb,Zn and As; 6) 1:500000 geological map all data are geometrically co registered to constitute integrated image sets .The pixel size of images is 50X50 M except TM image which has the pixel size resample into 25X25M.

  2. Remote Sensing Image Processing
    Various Spectral enhancements are applied to TM and MSS images such as contrast scaling band ratio linear transformation convolution and so on. Processed images are used to interpret rock stratigraphic units lineaments and circular features .Results show that TM color composite images especially 4(R) 5/2 (G) 4/3 (B) and 4 (R) 5/4(G) 5/2 (B) are effective on discerning rock properties and alteration belts. linear transformations such as K-L and slant transform are mainly used to enhance structure information Boundaries of main rock stratigraphic units can be correctly delineated .Most liniments reflect the fracturing while circular features are mainly the representations of circular textural and or color anomalies caused by magmatism and alteration.

  3. Textural Analysis
    Two textural analysis methods spatial gray tone dependence and Fourier Power Spectrum are adopted and programmed .Five textural features are extracted in the first method contrast inverse difference moments d correlation e energy homogeneity which indicates such image properties as gray level homogeneity textural coarseness etc. four directional measures and three frequent measures are calculated in FPS method to examine textural features of E-W S-N NE-SW NW-SE direction frequencies and high middle and low frequencies .A textural image of gray correlation is show. A northwest striking fault F16 remarkably demonstrated on the image also displayed is a east west striking F2 fracture zone much better enhanced than original image shows a high frequent image of MSS except those locating on the banks of Yangtze River most high value areas on the image distribute at similar sites as major intrusives and ore fields .One possible explanation of this phenomenon is that textural features on intrusive bodies and ore fields become relatively complicated because of intensive tectonomagmatism fracturing mineralization and alteration of country rock etc .This indicates that high frequent textural image may give certain hint on magnetism and mineralization.

  4. Quantitative Analysis of lineament (QAL)
    Several techniques including interactive convolution filtering window searching and line fitting split mask matching and Hough transform are used to detect lineaments on remote sensing image .various measures such as density frequency central symmetry spatial corridor average orientation are computed for detected lineaments .Cross points histogram rose gram in whole area or sub areas can be plotted on displaying device further data processing technique are also introduced and developed including entropy analysis ratio spectral analysis etc. on the resultant images local or regional features deep or surface information can be easily distinguished is a trend surface image of lineament density interpreted from color infrared air photo east west trending of lineament field is clearly manifested which reveals the features of deep fractures in the study area.

  5. Image Processing of Aeromag and Gravity Data.
    Several image processing techniques such as contrast scaling convolution are applied aero mag and gravity data to enhance details and anomalies .A convolution image of aeromag is given in areas with great changes of magnetics are clearly revealed most of which corresponded in the location of intrusive or mineral fields . Textural analysis technique is also used to process aeromag and gravity images .A high frequent image of aeromag data is shown on which high value areas indicates the existence of large intrusive bodies frequent image of gravity data which displays much enhanced lineaments of N-S direction.

    To integrate two different geophysical information A new technique magnetic gravitational correlation matrix analysis is proposed fore given moving window a matrix B is constructed by defining its elements B as the accurate frequency that aeromag value equals to 1 while gravity to j in the window .from matrix B five measures are calculated and displayed in image from 1 gravitational low magnetical low stressing gravitational high magnetical high stressing gravitational high magnetical low stressing low gravity magnetic correlation these measures are weighted sums of matrix B that synthetically demonstrate the information of two data sets and their relationship in spatial distribution A HGHM image is given in fig 1 which obviously shows the GHMH combination of two famous ore fields besides the combined anomalies of two data famous ore fields distribution feature is also reflected in the image The E W striking and NE SW striking structure indicated by aeromag data and gravity data respectively are simultaneously displayed An evidence of structural control over magmatism is revealed by the NE SW striking of Xihu intrusive which is in accordance with the striking of gravitational expressed structure.

Fig. 1 Resultant images


Interpretation and analysis from integrated of fracture structure
The structure outline interpreted from integrated image sets is shown in fig 2.


Fig. 2 Interpreted structure outline


  1. East west Striking Fracture Zone
    F2 fracture is the most important teconimagmatic zone that controls magma intrusion and mineralization .it is perfectly delineated on GHMH image while on remote sensing image it is some what obscure only appearing as small discontinuous lineaments. its much enhanced demonstration is obtained on textural images .these properties indicate that F2 fracture is a deep regional fracture most known mineral districts are located on this zone and huge EW faults compressed structure can be seen near the zone in field besides F2 F1 F3 are also major E-W fracture in the area.

  2. North south striking fracture zone
    N-S fracture zones are mainly interpreted from processed aeromag gravity and textural images except f8 which is remarkably appeared on remote sensing image then display on aeromag image as N-S trending anomaly zones and gradient zones and as obvious lineaments on power spectrum and MGCMA images and textual images of MSS and TM fractures with such image characteristics are mainly polyactive deep fractures. Their intersections with E-W fracture zones usually control the distribution of ore fields.

  3. North east striking and north west striking fractures
    A great number of NE-SW fractures exist in the area including striking faults accompanying with NE fold and larger regional fractures. The former can be interpreted from TM or MSS images as lineaments and the latter are usually shown on aeromag and gravity images as well as textural images for example blind fracture F21 being obscure in TM images is revealed on GHMH image .it constitutes thye north west boundary of Xihu intrusive and then can be deduced as a magma controlling fracture similar to NE SW striking ones NW SW fractures can also be classified into two kinds the first kind is cross fault and the second one is regional large fractures which cut through entire NE folding group. the largest fracture F16 is shown on TM images as obvious rock sliding on mountain area .it become relatively vague in Quanternary area and is remarkably enhanced on textual image on aeromag image it main set as high anomaly and gradient zone there are many small intrusive along the south east part of the fracture it is inferred that F16 is a fracture the provide passage way and room for magma tic intrusion.

  4. Fracture Structure Analysis
    The interpretation of integrated image sets shows that most of the EW and NS direction fraction in the area have evident appearances in magnetic and gravitational fields and display as discontinuous lineaments or lineament group on MSS and TM images intensive magmatism often occur along these fractures these phenomena indicate that EW and NS fractures are early formed deep fracture shows that the shape folding group in study area develops between EW fracture belts and that fold axes turn from NE direction to NEE and SWW direction when closing to fractures. This makes clear that the folding group is restricted by fractures and it can then be concluded that EW fractures had formed before indosinian period when NE folding occurred one of the most important NS fractures is F8 it only provides the passage way of magma intrusion but also results in great difference inlithofacies and thick ness between strata in its two sides The NS and EW fractures macros scopically control the tectonic development magmatism and mineralization.

    A great number of NW SW and NW SE striking faults have been mapped in field's geological work before. They are usually the strike faults and cross accompanying NE indosinian folds interpretation of interpretation of integrated image sets shows that there still exist regional large fractures of NE SW and NW SE striking which cause anomalies in geophysical fields they also act as the conduits of magma intrusion and become favorable sites for mineralization when they interest with EW and NS striking fractures.
Mineral predicting
  1. Variables
    Square grid each having an area of 500X500 M2 is chosen as prediction unit four kinds of variance are used in each unit remote sensing geological geophysical geochemical geological variables include emerging areas in each predicting unit of strata and intrusive rock which are favorable for mineralization textural measures of GTD and FPS length and cross point number of linear circular features are used as remote sensing variables as far geophysical variable besides original measuring value results of textural analysis and MGCMA are also used the highest anomaly value is taken unit variable when geochemical data is concerned.

  2. Characteristics space transformation
    let G be number of group Ng number of training units and µg the mean vector in the g group Cg the with in group divergence µo the mean vector of all G groups N the total training unit number.

    Then the Mahalanobis distance from mg to mo is ( Xuan Guorong )

    Dg (mg - mo )’ Cg-1 (mg- mo )

    the mean M distance from all mg- mo is 
    
    G
D = SagDg	   ag = Ng / N
   g=1
    D = tr [SagCg-1(mg- mo) (mg- mo)' ] = tr [ SagCg-1 Bg ]

    Let k = SagCg-1 Bg then D = tr(K)

    An orthogonal matrix A is used to trans form the original variables to maximize the M distance D*:

    D* = Tr[A ' KA]max A ' A = 1

    It is proved that Maximum D 8 can b e obtained when characteristic vectors of matrix [K+K] is used to constitutes the matrix A.

  3. Results
    The above mentioned features selection method is applied to the original multi source variables and discriminating classification is sub sequent used to the derived variables .to examine the effectiveness of different kinds of variables several classifications using different variable combinations have been implemented a part from that using alkl variables results show that the combination of geological and geochemical variables is most effective in discerning with or without mineral occurrences which create 93 percent corrections in training area textural area variables of remote sensing in combination with linear and circular feature variables produce a coorecteances ratio of 75 to 80 percent it reaches 78 to 83 percent when geochemical variables are addded in to textural ones mineral prognostic map of study area is made by weighed sum of all classification Most known copper deposits fall in to care ful score areas of the map five previously un known promising areas are pointed out after care full check with geological and geochemical data it is concluded that these prospective areas are all at the favorable sites of structure stratum rock property magmatic rock and geochemical anomaly.
Conclusion
The study reported shows that integration of remote sensing and multi source geodata gives a powerful tool for geo structure study and mineral prospecting and that application of image processing technique to geo data has its distinctive advantage remote sensing spatial information such as texture lineament and circular feature provides great effectives both in basic geological research and mineral deposit targeting the introducing of textual analysis technique in to geophysical data is meaniful it greatly enhances the structural information contained in original data the MGCMA method proposed in the study prove to be efficacious and over all distribution features of two data sets characteristic space analysis and sub sequent classification used in this paper turn to be successful results show that procedures and programs of mineral predicting adopted in the study are practicable.

Acknowledgements
The authors wish to think Prof. Zhung Perian and Prof. Feng Maosen of China university of Geosciences for their help and instruction to the project Mr Yao guoqing of the same university for developing the programs if lineaments detecting.

References
  • Robert M Haralick " Statistical and structural approaches to texture " Proc. IEEE Vol 67, PP786-804, 1979
  • Xuan Guorong, "Feature Selection of Mahalanobis Distance in Optimum Classification " Proc. of National Artificial Intelligence & pattern recognition vol.3. 1986