NEW METHODS FOR CLINKER PHASE RECOGNITION USING AUTOMATIC IMAGE ANALYSIS

摘要

Microscopic analysis provides information about the dimensions and the crystallographic composition of polished sections of Portland cement clinker. To reduce the amount of expert time needed to acquire reliable data and to make the work easier, we propose the use of modern tools offered by computer image analysis. The objective of this study is to develop a new automatic image analysis procedure for performing phase distribution analysis. New algorithms in digital image analysis of clinker are proposed and their advantages over the previous studies are shown. To complete the analysis by crystal measurement, a fast interactive method is developed. Our automatic image analyzing system uses color images which are acquired from the optical microscope by a video-camera and a PC. The image analysis procedure is based on the model of the well-known point counting process. Image points are taken in regular distances and they are assigned to the different crystal classes. A point is classified by the texture and color characteristics of its neighborhood. This allows to robustly identify the crystals in clinkers of different origins, with variations on the chemical etching. The tool for crystal analysis was designed to minimize human intervention and the burden of manual operations. The practical use shows the ease of use and the economy of this tool. The results agree to a great extent with those obtained by an expert via point counting. The tool will be extended to automatically work out morphological and structural parameters.