26 
the best gaussian functions are obtained and then the threhsolid 
is detected as the minimum error point when a pixel is assigned 
toa class; 
M2: in an interactive approach, by using graphic-pictorial devi 
ces for checking the selected threshold value. 
6,2 Thinning Procedure 
The result of P4 procedure is a binary image containing lines, 
which correspond to the reticle signs, and which have a varia- 
-ble width greater than one pixel. 
In order to obtain lines which have a constant width of just one 
pixel, so that the detection of nodes is simpler, faster and mo 
re exact, the P5 procedure must be executed. 
The P5 procedure returns a matrix, containing the medial axis of 
the lines, which is dimensionally equal to the inputone. The thin 
ning algorithm is then a transform operator, which does not topo 
logically change the connectivity properties of the image, and at 
the same time, controls the width and the position of the comput 
ed lines. 
The basic concept for the thimning process consists in removing, 
at subsequent iterations, certain points belonging to the bounda 
ry of the lines without modifying their connectivity. 
The digital image is scanned top-down and left-right, during sub 
sequent explorations, and the non-essential points are removed 
by extimating all the possible configurations between the examin 
ed pixel, centered to a 3x3 moving window, and its 8-neighbours. 
7. Conclusions 
One test image has been digitized with the device and has been 
processed with the previously described preprocessing procedures. 
In this experiment, instead of digitizing a photogram of the sce 
ne, we acquired the reticle itself. The reticle has lines whose 
width is.0.2 mm and spacing is 1. mm; the 512x512 digitized ma- 
trix has been processed by using firstly the P4 and then the P5 
procedures. (5) 
In fig.7 the input matrix is overlapped to the thinned matrix: 
we can see that the thinned reticle is entirely inside the width 
Of the lines of the digitized reticle,'so that the error in de- 
tecting the plane coordinates is quite defined by the resolu- 
tion : of the scanning device. In fig.8 a row oÉË the Fig.’ matrix 
is drawn: we can see that the thinned reticle (points with 0 va 
lue) is. well centered in the thick lines of the input matrix. 
5 4 1 
If a smaller error is needed, a greater resolution device has to