Sovremennye problemy distantsionnogo zondirovaniya Zemli iz kosmosa, 2009. Issue.6. Vol.1. P. 105-112
Benchmark of points neighborhood spectral representations for
points correspondence in different images
В.А. Гришин
, Е.В. Белинская
Институт космических исследований РАН, 117997 Москва, ул. Профсоюзная, 84/32
There are considered spectral representations for points neighborhoods. These representations are intended
for using in computer vision systems of flying apparatus. The key parameters of computer vision
systems heavily depend from accuracy and reliability of correspondence establishing of the same points
images of observed surface in different images. Computational complexity of computer vision algorithms
is determined by computational complexity of correspondence process. The method of points neighborhoods
images representation have considerable influence to characteristic of correspondence establish
process. Benchmark of different spectral representations for points neighborhood is presented in this article.
The principal parameters, which are considered, are the probability of correspondence anomalous
errors and computational complexity of correspondence process.
Keywords: computer vision systems of flying apparatus, correspondence algorithms, spectral representation of points neighborhoods
Full text