ind
sed
36]
the
nd
nal
ect
ed
he
on
rst
properly by introducing the weight :
where © is the standard deviation and k a proportionality
constant.
Then the least squares criterion becomes:
OQ = wi up +waotug+....... +WN*uN = min
where u; are observation residuals ,
and the solution vector is given by :
X - (A*W*A) I«(At*W*F) (4)
The standard deviations of the ( x;,y:) are the elements of :
(A W*A)!
where À is a N by 1 matrix of the coefficients of the weights of
the observations, W a diagonal weight matrix, and F an N by 1
matrix of constants indicating linearity of the condition
equations.
Equation (4) is iterated until the desired accuracy is achieved. In
any iteration the initial value of X was taken from the previous
iteration.
Experiments have been carried out for the above described
approach to evaluate the performance of the HT under various
conditions such as: variation of angle of lines, length of line
segments, number of lines in the target, noise contamination,
and quantization of the parameter space. It is found that for a
noisy image the best accuracy is given by four equally spaced
lines with normal distance quantization equal to one, and angle
quantization equal to one degree.
The main advantage of the proposed method for target detection
and accurate measurement is that the whole procedure is totally
automated. The target detection can be performed quickly and
safely and then the center of the circular target can be found in
the image either as the center of the corresponding ellipse-circle
or as the intersection of four equally spaced diameters of the
circle target.
The method works well even under the most noisy environment.
No limitations have been observed. Practical experiments
showed that the standard deviation of the target center position
in repetitive image acquisitions with different conditions of
illumination was in the subpixel level.
5. CONCLUSIONS
A target detection method based on the Hough Transform and
utilizing a white circle with four diagonal black lines in a black
background is proposed. The method provides automated and
safe target detection even in the most inconvenient images
(noisy,interrupted). It works in an gradient image and the
accuracy of the target detection is of subpixel magnitude.
The purpose of the recent paper is to underline to the
photogrammetrists the existence of a powerful tool named
Hough Transform which in a digital image environment can
offer a lot in the pattern recognition process. Except for the
253
described target detection application of the Hough Transform,
work is currently underway by the authors in applying HT in
image matching techniques.
REFERENCES
Arbuschi, G., Cantoni, V., Musso, G., 1983. Recognition and
location of mechanical parts using Hough transform. Image
Analysis, Pitman, London.
Ballard, D.H., Sklansky, J., 1976. A Ladder Structured
Decision Tree for Recognizing Tumors in Chest Radiographs.
IEEE Transaction on Computers, 25: 503-513.
Ballard, D.H., 1981. Generalizing the Hough Transform to
Detect Arbitrary Shapes. Pattern Recognition, 13(2): 111-122.
Ballard, D.H.,Brown, C.M., 1982. Computer Vision. Prentice-
Hall Inc., Englewood Cliffs, New Jersey, pp. 123-131.
Ballard, D.H., Kimball, O.A., 1983. Rigid body motion from
depth and optical flow. Computer Vision ,Graphics and Image
Processing, 22: 95-115.
Duba, R.D., Hart, P.E., 1972. Use of the Hough Transform to
Detect Lines and Curves in Pictures. Communication of the
ACM, 15(1): 11-15.
Fennema, C.L., Thompson, W.B., 1979. Velocity
determination in scenes containing several moving objects.
Computer Graphics and Image Processing, 9: 301-315.
Gonzalez, R.C., Wintz, P., 1987. Digital Image Processing.
Addison-Wesley.
Harjoko, A., 1990. An Optimal Target Design for The Hough
Transform. M.Sc Thesis, University of New Brunswick,
Canada.
Hederson, T.C., Fai, W.S., 1984. The 3D Hough Space
Transform. Pattern Recognition Letters, 2: 235-238.
Hough, P.V.C., 1962. Method and Means for Recognizing
Complex Patterns. U.S.Patent No.3069654, December 18.
Illingworth, J., Kittler, J., 1988. A survey of the Hough
Transform. Computer Vision,Graphics and Image Processing.
44: 87-116.
Inigo, R.M., McVey, E.S., Berger, B.J., Wirtz, M.J., 1984.
Machine vision applied to vehicle guidance. IEEE Transactions
on Pattern Analysis and Machine Intelligence, 6: 820-826.
Kushnir, M., Abe, K., Matsumoto, K., 1983. An Application
of the Hough Transform to the Recognition of printed Hebrew
Character. Pattern Recognition, 16:183-191.
Lutofo, R.A., Dangless, E.L., Mildford, D.J., Morgan, A.D.,
Morrtissey, F.T., Thomas, B.T., 1989. Hough Transform for
Transputer Arrays. In: 3rd International Conference on Image
Processing and its Applications, University of Warwick, U.K,
pp. 122-130.
Mikhail, E.M., Akey, M.L., Mitchell, O.R., 1984. Detection
and Subpixel Location of Photogrammetric Targets in Digital
Images. Photogrammetria, 39: 63-83.
Nixon, M., 1985. Application of the Hough Transform to
correct for linear variation of background illumination in images.
Pattern Recognition Letters, 3: 191-194.
Sayed, A.N., Paderas, F.C., Mikhail, E.M., 1989. Image
Matching and Target Location Using Digital Images. In:
ASPRS-ACSM Fall Convention, Cleveland, Ohio - U.S.A,
Technical Papers, pp. 482-492.