Quantitative assessment of laser-dazzling effects on a CCD-camera through pattern-recognition-algorithms performance measurements

We used pattern-recognition-algorithms performance as a measurement standard for laser-dazzled images. A black and white CCD-camera observed a scene containing different geometrical patterns, which had to be recognized by the algorithm. The camera was dazzled by a nanosecond frequency doubled Nd:YAG laser. Dazzling conditions were variable in laser repetition rate, pulse energy, geometrical forms size and position relative to the laser spot. We implemented algorithms based on edge detection, which locate areas with similar forms compared with a reference symbol, using either a degree of correlation assessment or a Fourier descriptors quantitative analysis. We also characterized the dazzled area size in the image. Thanks to a cross analysis of both criteria, we succeeded in quantitatively assessing the influence of laser-dazzling on the performances of the algorithms. We point out the key role of the effective distance between a geometrical form and the dazzled area in the image on the computed degree of correlation or Fourier descriptor value of this form. The analysis of these quantitative results contributes to the better understanding of laser-dazzling, which can be useful to design efficient means to protect imaging systems.