Snow particle automatic classification with texture operators

The backscattered data recorded by the meteorological radar is exploited for rainfall/snowfall rate calculation according to the Z-R relation. This relation is governed by parameters, which are influenced by the size and shape of the falling particles. The variety of snowflake types as well as the in class shape and size differences make this problem very difficult. In this paper the problem of automatic snow particle classification into snowflake and graupel with texture operators is addressed. Images of snow particles are recorded by the imaging system. For classification purposes the kNN and SVM techniques have been applied. There are compared many well known texture operators techniques: first order features, co-occurence matrices, run length codes, grey-tone difference matrix, local binary patterns, and also new approches for texture operators deffinition are introduced. The results gathered for the snow particle database, which images have not been normalized in regard to scale and rotation, proved to allow correct classification of 87% of data. Describing each image by two different texture operators improved the results to 89%. In case of classification for natural phenomena the performance of this system is satisfactory.