Предварительная обработка и классификация изображений мелкоструктурных объектов

Abstract

The main stages of processing and classification of images of fine-structure objects on the example of the percentage determination of different materials grains that are part of the cement clinker. The method of solving this problem is scalable for processing, pattern recognition and classification of images of nano-objects. To solve this problem the block diagram of the calibration (training) of the classifier is worked out in this work. The proposed approach is robust to the presence of fictitious and redundant features, it enables faster classification compared to most other methods. The study of spectral characteristics of the spatial interference showed quite a significant correlation of fluctuations of the neighboring pixels values.