3.1.2 Co-ordination of the reference control frame must
be done prior the survey. The reference control frame defines
the co-ordinate system for the stope and must thus be defined
to the highest degree of accuracy. The following steps are
taken in calculating the object space co-ordinates of the
centres of the circular retro-reflective targets on the reference
frame:
- Image Capture
- Distance Measurement
- Target Location and Identification
- Free Bundle Adjustment
The pre-calibrated camera is used to capture a series of
images of the reference frame from differing perspectives. In
addition to the images, the precise measurement of at least
one distance between extreme target centres is required. This
distance will provide the constraint of scale on the
photogrammetric determination of the co-ordinates of the
target centres. The precision of the distance measurement
carried out with a micrometer was better than 0.5mm. This
was sufficient in spite of the transfer of the scale from the
small frame to the 10m rock face of the study case. It should
be noted that the required accuracy refers to relative
displacement in individual portions of the wall, as opposed to
an overall absolute accuracy.
Target centres are then located by an image analysis process
which calculates the centre-of-gravity of the pixel grey levels
in a window defined around the target by a thresholding
process. Identification of the target, i.e. the association of the
targets with the correct number in the co-ordinate list, is semi-
automated. An initial eight targets are manually identified,
after which an automated identification process relying on a
direct linear transformation (DLT) resection calculation
locates the remaining control frame targets. This process
relies on a list containing the co-ordinates of all control points,
where initial estimates of their object space co-ordinates to the
nearest centimetre are sufficient.
The final co-ordinates of the targets are calculated by means
of a free bundle adjustment. Initial approximation of the
exterior orientation parameters (perspective centre and
rotation angles) are provided by the DLT calculation for the
identification of the targets, as described above. The pre-
calibrated camera parameters (principal distance, principal
point and lens distortion parameters) are held fixed in the
bundle solution.
3.2 Image Capture
Once in the stope, the reference co-ordinate system must be
established and the camera prepared for use, before images
can be captured.
As the environment of the stope is extremely hot and humid,
an on-site acclimatization period is required for the
equipment. In practise it was found that at least half and hour
exposure to stope conditions was needed to free the camera
lens and all electronic components of condensation
The reference co-ordinate system is established by attaching
the reference control frame to a suitable support pack (Figure
730
1), at one end ofthe rock face to be measured. To achieve the
geometric stability required for sub-millimetre determination
of the rock face digital elevation model, a high degree of
redundant images were captured. To cover the 10m long and
1m high rock face under investigation, a total of 60 images
was taken, including those necessary for the transfer from the
reference frame to the wall. In this way each surface point
was covered by an average of 6 images. The pre-analysis
showed that a sufficiently strong geometry for the overall
image formation could only be guaranteed by a set of control
points in front of the rock face. This was realised by
introducing target poles with circular targets, evenly
distributed in an area covering a band of approximately 1m
ahead of the face. These points only functioned as tie points
and could therefore be placed independently for the pre- and
post-blast photography.
4. DATA PROCESSING
Once the images have been captured, both before and after
the pre-conditioning blast event, the following steps are
needed to extract the relevant rock surface co-ordinates for
the deformation analysis.
- Camera Position Location
- Feature Extraction
- Image Matching
- Space Intersection
4.1 Camera Position Location
Before the features on the rock surface can be analysed, it is
necessary to determine the location and orientation of the
camera setups in object space, as defined by the reference co-
ordinate system. These parameters, the exterior orientation
parameters, w/x/¢ (rotation angles about the X/Y/Z axes)
and X/Y JZ. (perspective centre object space co-ordinates of
the camera's position) are determined by means of direct
linear transformations followed by bundle adjustment
calculations for each of the camera stations in turn.
4.1.1 Direct Linear Transformation (DLT): The initial
estimation of the camera exterior orientation parameters and
tie point positions are determined by means of a sequential
"folding-out, folding-in" DLT model. In an initial step, at least
six reference frame points are needed to determine the first
camera position. The DLT employed for this stage provides
exterior orientation parameters as well as the DLT
transformation co-efficients. These transformation co-
efficients are then used to evaluate the tie point positions
which provide reference points for the next camera position.
This resection-intersection sequence was originally intended
to be executed throughout the whole model, until
approximation values for all camera position and orientations
are determined for a subsequent complete bundle adjustment.
However, in practise, it proved necessary even at this
preparatory stage to execute sub-block bundle adjustments for
due to excessive error propagation.
4.1.2 Bundle Adjustment: The complete bundle adjustment
includes all of the data in a combined adjustment and allows
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996
