During last decades grows the interest of investigators to the methods of using the third level features for increasing identification accuracy in automated fingerprint recognition systems. It was noticed that the poroscopical maps carry important information about fingerprint pores, namely, their number, sizes, coordinates etc. However, at creation of automated identification technique, there arises a problem in connection with impossibility of using directly the pixel-by-pixel comparison methods of corresponding images. In this paper, we propose technique for comparison of poroscopical maps, based on using investigated earlier structural proximity assessment measure, which was determined using the gradient field of images. The poroscopical map is determined using some known algorithms of image processing, namely binarization, segmentation etc. The results of poroscopical maps processing and comparative analysis of some items from the database of Hong Kong Polytechnical Institute are given. It is shown that the proposed technique for poroscopical maps processing can be used in AFRS for increasing the accuracy of fingerprint identification.