Algorithm for clustering by means of assessment of local modes of a mixed distribution
N. V. Kolodnikova, Konstantin T. Protasov · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2003
In analysis of multispectral aerospace images of the Earth's surface and clouds with lacking a priori data, when an operator cannot select learning samples for pattern recognition, the only way for data structurization is the use of cluster analysis algorithms. A rather large number of versatile cluster analysis algorithms are developed by now, but particular data require an adaptive approach to analysis of particular images. We propose several cluster analysis algorithms that are mathematically justified with allowance for the specificity of analyzed NOAA AVHRR video data. Some examples of analysis of AVHRR images by the algorithms developed are presented.