CAM 1
CAM 2 CAM 4
CAM 3 CAM 5
TARGETS
SUBJECT
Figure 10. The position of the camera stations.
Following the recognition, location, and labelling
process the coordinates of the targets from each camera
station were passed to a software package called the
General Adjustment Program (GAP) developed at City
University, and the resulting 3-D data were used to
construct a wire frame view of the turbine blade surface,
see Figure 11.
Figure 11. A 3-D view of a section of the turbine.
The research indicates that in some situations where
there is prior knowledge of the subject to be measured
automatic target recognition is possible. Furthermore,
the use of subpixel techniques can enhance the accuracy
of CCD cameras beyond the nominal accuracy
suggested by sensor size and numbers of pixel. Hence,
such image processing hardware and software methods
may overcome some of the disadvantages of manual
target identification and location which is both time
consuming and tedious.
The investigations described in this paper used an
essentially simple structure which allowed the use of
straightforward strategies to achieve good results.
However, in most cases, more complex structures with
non-ideal imaging characteristics are required to be
measured. This research has allowed the authors to
become acquainted with some photogrammetric
methodology and has given insight into directions for
further work. Such work should address four main
areas:
(i) The target recognition, location and labelling
method could be seamlessly integrated into the
bundle adjustment method.
(ii) The use of transformations based on central
perspective projection e.g. epipolar, to extend the
labelling process beyond 2-D object assumptions.
(iii) The target location algorithms may be further
refined to give greater accuracy, reliability, or
efficiency. |
(iv) The binary method could be dispensed with if
alternative methods could be devised for use in grey
scale images.
Further research which will build upon the
multi-disciplinary approach is underway. It is hoped
that a greater understanding of photogrammetric
methodologies will result in a useful contribution to
machine vision problems.
6. ACKNOWLEDGEMENTS
The authors would like to acknowledge the support of
a number of colleagues: Prof. K.T.V. Grattan, Dr. S.
Robson, Dr. T J. Ellis, and the provision of the software
package "GAP" which was developed in the
Photogrammetry Unit in the Civil Engineering Dept,
City University.
7. REFERENCES
Bethel, D.J., 1990. Digital image processing in
photogrammetry, Photgrammetric record, B(76)
October, pp. 493-505.
Beyer, H.A., 1990, Linejitter and geometric calibration
of CCD cameras, ISPRS Journal of Photogrammetry
and Remote sensing, Vol 45, pp. 17-32.
Bhanu, B., 1986, Automatic Target Recognition : state
ofthe art survey, IEEE Transactions on Aerospace and
Electronic Systems, Vol AES 22, No 4, pp. 364-379.
Deng, T., 1987, Accuracy of position estimation by
centroid, Intelligent Robots and Computer Vision,
SPIE Vol 848, pp. 141-150.
Granshaw, S.L, 1980, Bundle adjustment methods in
Engineering Photogrammetry, Photogrammetric
record, 10(56), pp. 181-207.
Gruen, A.W., 1989. Digital Photogrammetric
Processing Systems: Current Status and Prospects,
Photogrammetric Engineering and Remote Sensing,
Vol.55, No.5.,pp.581-586.
Haralick, R.M. & Shapiro, L.G., 1985, Survey : Image
segmentation techniques, Computer Vision, Graphics,
and Image processing, Vol 29, pp. 100-132.
Mikhail, E.M., et al, 1984, Detection and subpixel
location of Photogrammetric targets in digital images,
Photogrammetria, Vol 39, pp. 63- 83.
Pavlidis, T., 1982, Algorithms for graphics and image
processing, Published by Springer-Verlag, Berlin, 416
pages.