Dalotto, Roque Sanchez
5 RESULTS
was used. It is a 35mm manual analogic camera, with Color-Ultron 1.8/50 lenses. Also,
35mm automatic analogic Yashica T4 camera was used, with Carl Zeiss T * Tessar 3.503%
objective.
As Digital camera, an Olympus C-820L Camera was used. Its main features are: digit
registration, 2Mb of memory for storage of images. Sensitive element composed Dy mai,
CCD of state solid of 8,5mm (1/3 of inch), 810000 pixels. It registers in 1024 X 768 pixels
White automatic balance. 5mm F2.8 Olympus Lens, with 5 elements in 4 groups.
4 METHOD
The program applies the "moving window" method based on iterative moving of a
exploration window or "kernel" (LILLESAND & KIEFER, 1987), defined by the user. The
correlation coefficient is calculated at each position.
With the largest correlation (or an interval of correlations), the central point of the window i
calculated, being written the image X, Y coordinates to an ASCII file. Other applications like
AutoCad or PhotoModeller can use them for more accurate tasks.
Program flow contemplates three stages: input, processing and output.
Data input can be made in three different ways: the first, by means of the routine of Digital
Correlation for files. In the first one, user indicates by means of the moving window, which wil
be the exploration kernel, contained in a raster file. It can be supplied to the program the
correlation threshold.
Second way needs numerical definition of the exploration kernel for its components or
elements. The data restriction is 400 elements (20 lines X 20 columns), and entered digital
values must be between 0 and 255.
Third way for input data consists in the definition of a window on the search image by means
of the mouse pointer, establishing a rectangular kernel for exploration.
The calculation method was centered average (PIECHEL, 1991) to determine the variance,
deviation, covariance and correlation factor. As data output, correlation results with X, Y
coordinates data of the points and calculated correlations are stored in text files
In marks applied over car structure were identified 10 of the 12 marks. Other sections of the
photo presented same patterns, with similar statistical conditions for digital correlation of
images.
162
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B5. Amsterdam 2000.