4. Coherent control channel
The local structure of SH is represented by diffraction grating patterns with sizes from 2 too 200um, periods in the range from 0.4 to 2.0um and different orientations. Illumination of such a patterns by coheerent light provides appearance of characteristic diffraction peaks in scattered light intensity distribution that can be captured by matrix photo-detector. Elements with different set of gratings will produce diffraction images witth different spatial distribution of diffraction peaks. To capture such an images an optical scheme represented on figuure 3 can be used in coherent control head construction.
The scheme uses laser module contained by four laser diodes: blue, green, red, and infraarred. Flexible fiberoptical light guides were used to deliver an illumination from all the lasers to optical head. TThe opened ends of optical fibers were mounted on visualising disc such that to provide the normal illumination of SSH surface. Such a construction allowed to visualise a 1st diffraction orders on a top surface of visualizer and captture of these bright points by photo detector array (PDA) while DC order was reflected back to visualizer and attenuuated by fiber ends mounting. The collimating objective refracts the emitted laser light such that the +/- 1st diffracction orders appear focused on lower surface of visualizer.
As the sizes of SH surface grating patterns are less than precision of optical head positioniing error, two ways of solving this problem were developed. The first one was the size of the laser spot on SH surfaacce was set up to be about 0.5mm and the second one was the application of CCF implementation concept in trainning of the control channel image recognition algorithms [Kumar (1992)]. For every chosen cell of SH surface with tthe size 0.5x0.5mm the set of training images was contained by the diffraction images of cell itself also the imagess of neighbour cells were added. This allowed to compensate the lack of precision of aiming channel positioningg and increase the stability of OES performance in presence of local stretches of SH surface. In order to increaase the stability of recognition algorithms in presence of possible imperfections of template examples of choseen SH, several SH examples must be used to gather the training data. Thus the set of training spectral images for eacch SH surface zone included the spectral images of zone itself, spectral images of neighbour zones taken fromm several template examples of chosen SH type.
Application of CCF implementation methods in image processing algorithms allowed ttoo obtain the mean discrimination error to be less than 5%, in some cases it was less than 1%.
4. Coherent control channel The local structure of SH is represented by diffraction grating patterns with sizes from 2 too 200um, periods in the range from 0.4 to 2.0um and different orientations. Illumination of such a patterns by coheerent light provides appearance of characteristic diffraction peaks in scattered light intensity distribution that can be captured by matrix photo-detector. Elements with different set of gratings will produce diffraction images witth different spatial distribution of diffraction peaks. To capture such an images an optical scheme represented on figuure 3 can be used in coherent control head construction. The scheme uses laser module contained by four laser diodes: blue, green, red, and infraarred. Flexible fiberoptical light guides were used to deliver an illumination from all the lasers to optical head. TThe opened ends of optical fibers were mounted on visualising disc such that to provide the normal illumination of SSH surface. Such a construction allowed to visualise a 1st diffraction orders on a top surface of visualizer and captture of these bright points by photo detector array (PDA) while DC order was reflected back to visualizer and attenuuated by fiber ends mounting. The collimating objective refracts the emitted laser light such that the +/- 1st diffracction orders appear focused on lower surface of visualizer.As the sizes of SH surface grating patterns are less than precision of optical head positioniing error, two ways of solving this problem were developed. The first one was the size of the laser spot on SH surfaacce was set up to be about 0.5mm and the second one was the application of CCF implementation concept in trainning of the control channel image recognition algorithms [Kumar (1992)]. For every chosen cell of SH surface with tthe size 0.5x0.5mm the set of training images was contained by the diffraction images of cell itself also the imagess of neighbour cells were added. This allowed to compensate the lack of precision of aiming channel positioningg and increase the stability of OES performance in presence of local stretches of SH surface. In order to increaase the stability of recognition algorithms in presence of possible imperfections of template examples of choseen SH, several SH examples must be used to gather the training data. Thus the set of training spectral images for eacch SH surface zone included the spectral images of zone itself, spectral images of neighbour zones taken fromm several template examples of chosen SH type. Application of CCF implementation methods in image processing algorithms allowed ttoo obtain the mean discrimination error to be less than 5%, in some cases it was less than 1%.
การแปล กรุณารอสักครู่..
