ASEG Extended Abstracts

ISSN: 2202-0586 (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/texg19

Application of automated detection techniques

in magnetic data for identification of Cu-Au
porphyries

Matthew Hope, Barry Bourne, Simon Crosato, Brendan Howe, Eun-Jung

Holden, Shih Ching Fu & Peter Kovesi

To cite this article: Matthew Hope, Barry Bourne, Simon Crosato, Brendan Howe, Eun-Jung
Holden, Shih Ching Fu & Peter Kovesi (2010) Application of automated detection techniques in
magnetic data for identification of Cu-Au porphyries, ASEG Extended Abstracts, 2010:1, 1-6, DOI:
10.1081/22020586.2010.12041956

To link to this article: https://doi.org/10.1081/22020586.2010.12041956

Published online: 12 Feb 2019.

Submit your article to this journal

Article views: 116

View related articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=texg19
Application of automated detection techniques in magnetic data for
identification of Cu-Au porphyries
Matthew Hope, Barry Bourne
Simon Crosato, Brendan Howe
Barrick Gold of Australia
Level 10, 2 Mill St, Perth,
W.A, 6000, Australia
mhope@barrick.com
SUMMARY
Automated shape recognition technology has been
developed for application to porphyry exploration,
through a joint research initiative between Barrick Gold
and the Centre of Exploration Targeting at the University
of Western Australia, commencing in 2009. The result is
the Development of the Porphyry Texture Filter for
application to magnetic datasets.
Many mineralised porphyries display concentric zonation
in their magnetic character as a by-product of extensive
hydrothermal alteration systems and secondary magnetite
development/destruction. This characteristic magnetic
signature can be exploited by image processing
techniques enabling the enhancement, identification and
quantification of features. Features must agree with a
user-defined set of criteria for size, shape and magnetic
contrast.
Development of the technique was carried out on the
world class Reko Diq porphyry system resulting in
successful identification of all major known mineralised
porphyry centres and additional targets within the camp.
User control over filter parameters has resulted in the
successful application of the filter on projects in a range
of geological and erosional environments.
The ability to rapidly characterise porphyry-like
signatures using mathematical principles and geometries
results in an unbiased geophysical target layer. When
integrated with other geoscientific data, the filter has
consistently supported target generation activities.
Examples of the Porphyry Texture Filter application and
results from Reko Diq, Grasberg and active exploration
projects are shown.
Key Words: Porphyry, Filter, Magnetics, (?),
Interactive processing, Automated shape recognition.
INTRODUCTION
Magnetic datasets are a common and inexpensive geophysical
tool for porphyry Cu-Au exploration used throughout the
exploration process. As porphyry exploration for undercover
systems gathers pace and the number of targets with classical
geophysical signatures dwindles, a need has arisen for flexible
and interactive processing tools. These tools should be
capable of producing meaningful results in a variety of target
regimes at all scales and geological environments.
ASEG 2010 - Sydney, Australia
Eun-Jung Holden, Shih Ching Fu
Peter Kovesi
Centre for Exploration Targeting
The University of Western Australia
35 Stirling Hwy, Crawley, W.A, 6009, Australia
(eunjung, sfu)@cyllene.uwa.edu.au
pk@csse.uwa.edu.au
Automated shape recognition technology has been developed
for porphyry copper-gold exploration, through a joint research
initiative between Barrick Gold and the Centre of Exploration
Targeting at the University of Western Australia (CET). The
result of this initiative is the Porphyry Texture Filter. The
filter is designed for magnetic datasets and uses characteristics
of magnetic anomalies related to porphyry style alteration and
mineralization. Outputs from the filter include definition of
targets via vector boundary files. These boundaries when
coupled with qualitative interpretation outputs, significantly
aid the exploration process.
The filter uses an interactive approach which requires
significant user definition. These input parameters require the
explorer to consider the specific nature of the porphyry target
and the expected magnetic signature. The filter outputs can
then be designed to enhance specific size, shape and magnetic
contrasts of anomalies whilst suppressing non-conforming
signatures.
The development and testing of the filter was carried out using
world-class porphyry systems such as Reko Diq and Grasberg
to build confidence in the results. The technique has since
been applied to active exploration programs with very
encouraging results.
THEORY & BACKGROUND
The porphyry texture filter utilises the diagnostic
characteristics of magnetic anomalies associated with Cu-Au
porphyry alteration-mineralization footprint to automatically
screen magnetic datasets. The filter seeks out magnetic
signatures consistent with user-defined input parameters via a
three stage process of circular feature detection, target
magnetic contrast and boundary detection. The resulting data
includes contrast-enhanced grid data and vector output for
both centres of radial symmetry and target boundary
detection. These outputs can be readily incorporated into
common exploration software packages for analysis.
Idealised Porphyry Magnetic Signature
Many Cu-Au porphyry systems display a magnetic anomaly
associated with generation of hydrothermal magnetite in the
potassic alteration cores (Clark et al., 1992). This zone is often
surrounded by a phyllic alteration zone where magnetite is
destroyed (Sillitoe, 2009). The resulting anomaly is often a
localised sub-circular magnetic high surrounded by a
magnetic low before returning to outboard background
magnetic intensity (Figure 1). This signature can be readily
recognised in many world-class deposits including Reko Diq,
1
Automated Detection Techniques for Porphyry Cu-Au Exploration
(Fletcher et al., 2009) Bajo del la Alumbrera, (Proffett, 1996)
and Grasberg (Meinert et al., 1997).
Figure 1. Magnetic profile anomaly of Cu-Au porphyry
(After Clark et al. 1992)
a) b)
Figure 2. H14 and H15 porphyry system, Reko Diq a)
outcrop alteration b) RTP magnetic image (After Hope,
2010)
Factors including, but not limited to, erosion level, alteration
overprinting, magnetic susceptibility of host rocks and post
emplacement deformation can significantly influence the
resulting magnetic anomaly and, in some cases, render them
difficult to recognise in magnetic data. These challenges are
specific to each exploration environment and a single filter
tool may not be applicable in all cases. This has led to the
development of the porphyry texture filter permitting full user
definition of target anomalism.
Porphyry Texture Filter Theory
The full theoretical and mathematical explanation of the
porphyry texture filter can be found in Holden et al. (2010). A
brief summary of this work is included herein.
The approach utilises a 3 stage process:
1) Circular feature detection utilising a radial
symmetry transform (RST) developed by Loy
and Zelinsky (2003)
2) Validation of centres of symmetry based on
determining the magnetic contrast with its
surroundings using the target magnetic contrast
transform (TMCT)
3) Determination of size and geometry of
confirmed targets using spline contours to
output a vector product
ASEG 2010 - Sydney, Australia
Hope et al.
Circular feature detection first seeks the centre of elevated
circular features by identifying gradient converge/divergence.
The RST allows the user to define the radii of interest and
degree of circularity. The final symmetry image is a
summation of all radii of symmetry at each point. A final
stage of non-maximal suppression of this image results in
definition of centres.
The application of the target magnetic contrast transform then
tests each centre identified by RST. Contrast is calculated by
examining the difference in values about each location at
different radii and combining these responses into a summary
statistic. A value is calculated for each location by summing
the difference in value at a given radius. This results in an
enhancement of circular features of interest whilst suppressing
anomalies which fail to meet user defined criteria for size,
geometry and contrast. Amplitudes of the TMCT are largest
on the edge of magnetic anomalies and create a halo around
the target (Figure 3).
a) Input Image b) TMCT output
Figure 3. Result of TMCT on synthetic data. Red box
highlights signatures consistent with typical porphyry
magnetic response (After Holden et al. 2010)
Finally a deformable spline approach suggested by Williams
and Shah (1990) is used to assign a boundary to each
validated centre. The spline converges onto the feature
periphery detected by the TMCT output subject to some
geometric constraints. The final shape thus defines the outer
edge of the anomalous magnetic zone often associated with
limit of hydrothermal magnetite in porphyry systems. This
boundary, in addition to the centre of each radius of
symmetry, is a vector form converting qualitative images into
quantitative results which conform to user-defined parameters.
In addition to the vector results, grid images of radial
symmetry transform and magnetic contrast are produced.
These images can be used to validate the final results and
ensure desired data is being captured by filter parameters. The
speed of filter operation allows a rapid iterative approach to
be applied whilst parameter selection is optimised for target
style. Figure 4 displays the stages of boundary selection using
a simple magnetic anomaly.
2
Automated Detection Techniques for Porphyry Cu-Au Exploration
Figure 4. Porphyry texture filter stages and outputs (After
Hope et al., 2009)
USER INPUT CONTROLS
The significant advantage of this interactive filter control over
traditional gradient filters such as analytic signal is the ability
to optimise the filter results specific to a set of exploration
criteria. Selection of the user-defined parameters can be
readily modified to incorporate geological context such as
target size, magnetic contrast with host rocks, structural
regime, and depth of emplacement or minimum distance
between two targets. This flexibility allows the filter to be
applied to data at different stages in the exploration pipeline
from early generative porphyry cluster definition to
camp/local scale targeting. The ability to rapidly produce a
suite of results also allows an iterative process and selection of
most appropriate results to be completed in a timely fashion.
Input Parameters
Search Radii Selection
This parameter is critical in determining initial results from
RST and is given as a minimum-maximum range defined by
number of cells in image. Appropriate selection should
consider target size range and magnetic data resolution.
Radii Calculation Method
The final assigned radius of symmetry for each point can be
calculated by either using the mean or maximum radius of
symmetry about each point.
Radial Strictness
The RST allows definition of target circularity from 1-3 (least
to most circular). Syn-post deformation of a target district
may cause elongation of targets and should be considered
prior to setting this value.
Point Suppression Threshold
This value controls which centres identified by the RST are
treated as potential porphyry targets. A range of thresholds
ASEG 2010 - Sydney, Australia
Hope et al.
can be delivered and rapidly verified. Selection of appropriate
value is critical and is heavily influenced by depth of
emplacement, alteration overprinting or magnetic host
Exclusion Radius
This parameter allows the user to determine the closest two
centres may be located. This is important to prevent excessive
merging of boundaries. In the case of two centres being
located within the exclusion radius, the most symmetrical
centre will be retained.
The number of defined parameters and flexibility of the filter
required close interaction between the CET image processing
group and Barrick geophysicists. To refine parameter
selection, appropriate datasets were supplied to the CET and
regular review of results and feedback sessions carried out.
This close one-on-one relationship between two disciplines
allowed the filter development to be highly dynamic and
ultimately led to the addition of the boundary outputs.
APPLICATION
Significant testing and verification of the process was required
to build confidence in the outputs prior to implementation on
active exploration projects. Testing focussed on successfully
replicating known porphyry centres whilst minimising the
number of false positives. This process allowed parameters to
be refined for specific targets and an iterative improvement in
application. The Reko Diq porphyry cluster was nominated as
a test dataset due to its large number of magnetic targets and
relatively simple geology. Additional testing was then made
on a number of world-class porphyry targets.
Once successful application had been demonstrated, the filter
was incorporated into active global exploration programs. It
has quickly been recognised that in new exploration
environments an existing known porphyry target is invaluable
for appropriate tuning of the filter parameters. Due to filter
flexibility, a very high success rate in defining known systems
has been achieved. This success rate and implementation of
the filter in increasingly challenging environments has built
confidence in the results. The filter outputs now form a
standard magnetic data layer for porphyry exploration across
Barrick global exploration and have been actively applied in
South America, Asia-Pacific and Eurasia.
Testing
Reko Diq
The Reko Diq project was chosen as an ideal test dataset for
this project due to the high quality magnetic data available,
relatively simple geology and large number of magnetic
anomalies. The Reko Diq porphyry cluster is located in the
Baluchistan province, Pakistan. A large number of variably
mineralised satellite porphyries are also noted in addition to
barren diorite phases.
The host sequence of low-susceptibility andesite and
volcaniclastics overlying a thick sedimentary package and
outcropping to sub-cropping potassic alteration zones allow
the porphyry systems to be readily recognised as strong
magnetic anomalies. However this is complicated by the
presence of strongly magnetic but barren, post-mineral
diorites that have similar amplitude and size.
3
Automated Detection Techniques for Porphyry Cu-Au Exploration
After assessing a number of results it was found that the filter
was able to identify all major known porphyry centres in this
environment whilst also rejecting barren phases in most cases.
The preferred solution (Figure 5) was selected on the basis
that all large systems were identified without excessive
number of false positives. The relatively low grade systems
necessitate a large target size and this parameter is the most
common reason some known porphyries failed to be
identified. This can be corrected in the filter parameters but
would dilute exploration focus into these smaller targets.
a)
b)
Figure 5. Reko Diq porphyry cluster a) RTP data with
known porphyry and boundary results b) Simplified
interpreted geology with boundary results (After Hope &
Fletcher, 2010)
A statistical assessment reveals that of 29 known porphyry
targets (not all shown), 21 were successfully identified.
Review of the 8 targets that failed to be identified show that
target size was the key failure parameter, some targets also
failed due to elongated geometry.
Grasberg
ASEG 2010 - Sydney, Australia
Hope et al.
The giant Grasberg/Ertsberg porphyry deposits are located in
the Ertsberg district of Irian Jaya and collectively represent
one of the largest Cu-Au resources in the world (Meinert et al.
1997). The intrusive centres and associated skarns are hosted
in a sedimentary sequence of limestone and sandstone with
low magnetic susceptibility. The low latitude of this system
presents a challenge to the use of reduction to pole magnetic
data and necessitates amplitude corrections or use of the
analytic signal for most exploration purposes. Data used for
this study was digitised from magnetic contours in Potter
(1996).
Due to the large size and significant contrast associated with
this system only a small number of results (4) were returned
from the filter that are consistent with target criteria. All four
targets correlate well with the mapped position of intrusive
centres at both Grasberg and Ertsberg. In addition a fourth
target to the south correlates with a small, largely covered
mapped intrusive suggestive of additional size potential under
cover. Figure 6 shows the results of the Porphyry Texture
filter application at Grasberg.
a)
b)
4
Automated Detection Techniques for Porphyry Cu-Au Exploration Hope et al.

c)
Figure 6. Grasberg-Ertsberg area a) Radial Symmetry
with located centres b) RTP Image with boundary results
(after Potter 1996) c) Geology of Grasberg area with a)
boundary results from porphyry filter (After Meinert et
al. 1997)

Exploration Applications
Based on various tests and case studies, this technique has
been applied to active exploration programs. Application
within Barrick on two porphyry exploration programs is
presented here to demonstrate integration of the filter results
with other datasets for targeting in differing geological
environments.

Project 1
This project area was selected on the basis of structural setting
and stream sediment anomalism, and has been covered by a b)
helicopter magnetic survey, which shows significant survey Figure 7. a) RTP image with filter results on local scale b)
related noise issues. The dominant geological units are weakly Geology and geochemistry with boundary results (After
magnetic basalts juxtaposed against schists. These have been Hope et al., 2009)
variously intruded by younger diorites and dacites.

The targets are moderate-sized shallow to outcropping
porphyry systems which have a strong-moderate magnetic
contrast with low susceptibility host basalts. Due to inherent
suppression of non-circular features in the filter the variation
in data across line does not significantly impact the results.
These results display the robust nature of the filter and allow
for the use of sub-optimal data in many cases with minimal
effect.
Close assessment of the results show two distinct target
groups detected by the porphyry texture filter (Figure 7):
a) Large highly magnetic fresh diorites, showing
significant remanant magnetism; and
b) Smaller subtle anomalies consistent with target
style dacites
When reviewed against field mapping the high ranking trend
consists of variably altered dacites with the filter
preferentially detecting the mineralised intrusive centres in the
wall rock basalts. Soil geochemistry shows a clear correlation
between located centres and gold anomalism in outcropping
areas. In addition to known porphyries several additional
targets located under shallow cover with signatures consistent
with target are noted (Figure 7).
This methodology of calibrating with known systems within a
project area and integrating results with other geological
datasets is the optimal application of porphyry texture filter
outputs.
Project 2
Located in a challenging porphyry exploration environment
this project is targeting buried porphyry systems. Magnetic
cores are anticipated to be >200m depth and significant
argillic alteration overprinting and magnetite destruction is
present. In addition depth constraints render economic target
size to be large. As such, weak magnetic anomalies with large
diffuse geometries are the most appropriate magnetic targets.
A further complication is the presence of highly magnetic
post-mineral diorite intrusions which dominate the magnetic
data. Ground magnetic data acquired at 50m station intervals
was used for processing.

ASEG 2010 - Sydney, Australia 5

Automated Detection Techniques for Porphyry Cu-Au Exploration
Figure 8. a) Ground magnetic data from Project 2 with
boundary results shown in black
A review of the results shown in Figure 8 at this project show
a number of boundaries coincident with existing prospects
defined by copper in soil anomalies and alteration. Drilling of
the large boundary result coincident with copper anomalies
has confirmed the presence of a deep magnetic potassic core.
Several magnetic features were discounted as related to post-
mineral intrusive following field checking. Results were
prioritised by size and coincidence with geochemical
anomalism
CONCLUSIONS
The Porphyry Texture Filter successfully identifies porphyry
system signature in magnetic data in a variety of geological
environments. The high level of flexibility and user control
allows geological understanding and target styles to be
considered to optimise the results on a project specific basis.
Correct parameter selection is critical to achieving effective
outputs and in ideal situations a known system should be used
to test parameter selection prior to application at larger scales.
The filter can be applied to exploration projects at a variety of
scales and is effective on data of variable quality due to built-
in rejection of non-circular geometries. The boundary file
outputs take a crucial leap in converting image products into
vector layers which attempt to quantify targets meeting preset
criteria. These layers can then be rapidly integrated into a GIS
environment for ranking and assessment.
The application of this magnetic filter is now standard
procedure within Barrick and adds an additional exploration
layer to the potential field data suite for porphyry exploration.
Barrick continues to work closely with the CET group to
further the application of the Porphyry Texture Filter.
ASEG 2010 - Sydney, Australia
Hope et al.
ACKNOWLEDGEMENTS
The authors thank Barrick Gold Corporation for allowing
presentation of data used in this paper.
REFERENCES
Clark, D. French, D. Lackie, M. and Schmidt, P., 1992,
Magnetic petrology: Application of intergrated rock magnetic
and petrological techniques to geological interpretation of
magnetic surveys, Exploration Geophysics, 23, 65-68.
Fletcher, T., Rehmann, A., Brockway, H., Livesey, T. and
Brommecker, R., 2009, The discovery and geology of the
giant Reko Diq copper-gold porphyry deposit, Chagai
Porphyry Copper Belt, Baluchistan, Pakistan, Proceedings of
the 2009 NewGenGold Conference, Perth 23-24 November
2009.
Holden, E., Fu, S., Kovesi, P., Dentith, M., Bourne, B., and
Hope, M., 2010, Automated Identification of Magnetic
Responses from Porphyry Systems, ASEG Conference 2010,
Extended Abstractcs, this volume
Hope, M., Crosato, S and Wallace, Y., 2009, Internal Barrick
Presentation, Exploration Conference, 2009, Unpublished.
Hope, M and Fletcher, T., 2009, Internal Barrick report,
Unpublished.
Hope, M., 2010, Internal Barrick Report, Unpublished.
Potter, D.R., 1996, What makes Grasberg anomalous,
implications for future exploration, Porphyry Related Copper
and Gold deposits of the Asia Pacific Region (Conference
Proceedings), Australian Mineral Foundation, 10.1-10.13.
Proffett J., 2003, Geology of the Bajo de la Alumbrera
Porphyry Copper-Gold Deposit, Argentina, Economoc
Geology vol. 98, 1535-1574.
Loy, G. and Zelinsky, A., 2003, Fast Radial Symmetry for
Detecting Points of Interest, IEEE Transaction on Pattern
Analysis and Machine Intelligence, 25(8), 959-973.
Meinert, L.D., Hefton, K.K., Mayes, D., and Tasiran, I. 1997,
Geology, Zonation, and Fluid Evolution of the Big Gossan
Cu-Au Skarn Deposit, Ertsberg District, Irian Jaya. Economic
Geology, vol. 92, no. 5, 509-534.
Sillitoe, R., 2009, Gold-rich Porphyry Copper Deposits:
Geological Model and Exploration Implications, Mineral
Deposit Modelling, Geological Association of Canada,
Special Paper 40, 465-478.
Williams, D. and Shah, M., 1990, A Fast Algorithm for Active
Contours, Proceedings of the Third International Conference
on Computer Vision, 592-595.
6