Changes between Version 6 and Version 7 of Tutorial/GeometricCalibration

Timestamp:

Aug 28, 2013, 2:57:45 PM (11 years ago)

Author:

sommeria

Comment:

--

Tutorial/GeometricCalibration

@@ -1 @@
-= Simple calibration =
+= Simple scaling =
 Open again a test image in 'UVMAT_DEMO01_pair'.(accessible on http://servforge.legi.grenoble-inp.fr/pub/soft-uvmat/)
 
@@ -13 +13 @@
 To see the calibration points on the image, first display the whole image by unselecting '''fix''' (tag '''[!CheckFixLimits]''') in the frame '''Coordinates''' of uvmat. Then press '''PLOT PTS''' in '''geometry_calib'''.
 
+To perform the calibration, press '''APPLY''', first with the default option 'rescale' in '''calib_type'''.  The image is now displayed in phys coordinates. 
+
+
+= Calibration with reference points =
+
+An alternative method of calibration consists of using a set of reference points whose physical coordinates are known. Open with '''uvmat''' an image in 'UVMAT_DEMO06_PIVconvection/Dalsa1' (accessible on http://servforge.legi.grenoble-inp.fr/pub/soft-uvmat/)
+
+Select in the menu bar '''Tools/geometric calibration'''. Mark the four box corners of the box with the mouse (left hand button). Their coordinates in pixels are displayed in the two last column of the table '''!ListCoord''' in the GUI '''geometry_calib'''. To clear the table for corrections push the button '''CLEAR_PTS''', or for a single line, use the key board backward arrow. To improve the position on the image, use the zoom and directional arrows. We find the coordinates of the four calibration points in pixels:
+
+(X,Y)=(80.3, 81.6), (982.3, 86.1), (978.9, 937.4), (71.2, 929.5).
+
+The corresponding physical coordinates are known to be
+
+(x,y)=(0,0),(58.8,0),(58.8,55.1),(0,55.1), with an origin (0,0) taken at the lower left (and z=0).
+
+Introduce those in the two first columns of the table [!ListCoord]. This can be conveniently done by copy-paste Matlab vector x=[0 58.8 58.8 0] in the upper line of the x column, and y=[0 0 55.1 55.1] in the y column (use carriage return to validate the input). 
+
+To perform the calibration, press '''APPLY''', first with the default option 'rescale' in '''calib_type'''. The image is now displayed in phys coordinates. We observe that the rectangular frame is slightly rotated. furthermore the displayed precision, about 3 pixels, is not excellent.
+
+To improve the precision we then apply the option 'linear' in '''calib_type''', which seeks a general linear transform, including rotation. The precision is indeed improved, about 1 pixel. The previous xml file has been saved with a ~ ('Dalsa1.xml~') so it can be reverted in case of error.