Abstract:

A micro-reservoir pore network pattern recognition method is designed by the repeated observation and experiments of a large number of the casting thin section of the rock sample. The key point of this method is that the transform from the RGB（red, green, blue） space into the YUV(Y-brightness, U, V-Color) space make the pore throat network  be identified into good clusterings, and then by use of the morphological watershed segmentation method, segment the pore and throat of the casting thin sections of rock samples. The identified pore-throat networks by this method and the general image segmentation techniques are comparative studied. The comparison results show that the pore-throat networks are more complete and clear by this method identifies. The surface porosity of casting thin section of the rock sample to be identified is calculated by use of Axiovision4.0 software belongs to the German Zeiss image processing system at the same time. Calculation results show that the recognition effect is obvious, relative error is smaller; the micro-pore network of reservoirs accuracy of pattern recognition is improved.

Key words: pore-throat network, pores connected region, morphological watershed, pattern recognition, reservoirs