An automatic classification method based on machine learning is proposed to distinguish between true and false laser-induced damage in large aperture optics. First, far-field light intensity distributions are calculated via numerical calculations based on both the finite-difference time-domain and the Fourier optical angle spectrum theory for Maxwell’s equations. The feature vectors are presented to describe the possible damage sites, which include true and false damage sites. Finally, a kernel-based extreme learning machine is used for automatic recognition of the true sites and false sites. The method studied in this paper achieves good recognition of false damage, which includes a variety of types, especially attachment-type false damage, which has rarely been studied before.
Based on the multiple pulses joining, the cascaded photodetection is experimentally researched on high-contrast measurement of ns high-power laser pulse. The ultrafast photodetectors with the saturation characteristics are used. A joining method for multi-pulse waveforms in the nonlinear region is put forward. The experimental results for ns step-pulse at SG-III laser system show the contrast of ~ 400:1 is achieved, in accord with designed contrast value.
Damage inspection of the large aperture components is required for Large, high-power laser systems. Dark-field
imaging technology is used to enhance resolution of defects. Because there are several of the optics which are laid with
Brewster angle in optics online inspection and the image collected by CCD includes many noises, so the image are quite
complex. A kind of image processing method is introduced, which is based on classical method about edge detection.
Gray restrain is used and the relations between the pixel and its eight neighbours are considered in calculating the
gradient. The defect size is measured and damage defect of optics is analyzed using the image processing methodology.
The new approach produces nice result.