Baklanova O.  

Development of algorithms for image recognition needed to assess the quality of the mineral species in the mining industry

Knowledge of the modal composition of the rock or ore is very important for solving the mineralogical and technological issues. Today, in the analysis of rocks are widely used both traditional and modern mineralogical and analytical techniques: optical and electron microscopy, X-ray diffraction, electron microprobe, and elemental analyzes, and the methods are widely involved studying the physical properties of minerals. With the expansion and deepening of knowledge about minerals are gaining the technical means to investigate their microstructure.
The purpose of this paper is development of new methods and algorithms for raster image processing, resulting microscopy to tackle problems of "machine" for evaluation of the quality of the mineral species in the mining industry.
This work is supposed to perform cluster analysis for segmentation of color images. The classic version of the cluster analysis, focused on a random selection of centroids unacceptable to adequately address the problems due to variations in the resulting picture segmentation, which, in turn, is highly dependent on the filing of observations at the input of the algorithm. As the results of processing a test image, each picture is different form of segmentation obtained segments. As the criteria for assessing the quality of the segmentation is expected to use a combined standard error, Hausdorff measure, the function of Liu and Yang. Also be used visual assessment, which exhibited a hundred point scale, based on the adequacy of the segmentation of objects in the problem of interest. It will be maximum if after segmentation image is divided into segments, perfectly inscribed in the border search objects.
These methods and algorithms might be used for raster image processing, resulting microscopy to tackle problems of "machine" for evaluation of the quality of the mineral species in the mining industry.