Image Analysis Of Cement Micro-morphology

In order to improve the research of the high-performance cement, in recent years, researchers have been probing to perform the innovation which combines the advanced information technology with the traditional experimental method on the research of new cement. It plays an important role in designing new cement to model the process of cement hydration through three-dimensional (3-D) simulation and reveal its inherent mechanism.However, there exist several drawbacks in the present 3-D simulation model of cement hydration. One is the parameters of the model are set only according to the initial condition of cement paste, which may result in the deviation from the reality in the simulation process of cement hydration. In order to solve the problem, we propose a new method to set up the 3-D simulation model: obtain some main hydrates' chronological variation by analyzing a large number of scanning electronic microscope (SEM) images of cement micro-morphology of different time. Based on the chronological variation, a reasonable 3-D model could be established. This thesis finished the basic part of the research work, that is:First of all, we obtained a large number of SEM images of cement micro-morphology of different time in process of cement hydration according to the standard experimental course. Then we developed software to manage the images. The common algorithms of image processing are integrated in the image database of cement micro-morphology. We expect the database become a promising expandable platform for the further work.The major part of my work is the analysis of images of cement micro-morphology and recognition of the regions of main hydrates' in the images. The images can be segmented into several sections, each of which is dominated by some specific features. Then pattern recognition can be realized through template matching. Image segmentation is very important for image analysis. Considering the abundant texture information contained in the images of cement micro-morphology, we extract the texture features by employing wavelet transform. In order to speed up the image processing and improve the segmentation, we made use of a novel algorithm combining fuzzy C-mean clustering with genetic algorithm. When applying the hybrid algorithm to some images, we acquired satisfactory segmentation of three main hydrates. Finally we calculated the relative proportions of these three hydrates.The methods of the texture analysis in the present thesis are effective for the subsequent work. Since we integrated many useful algorithms into the image database of cement micro-morphology, we wish it will be convenient and time-saving for users to analyze cement images. We expect the current methods be helpful in the further research.