Application of Wavelet Transform in the Image Analysis of Soil Microstructure

Abstract

Along with the rapid development of computer vision, image analysis technique was proposed to investigate engineering materials, such as soil [1, 2, 3]. First, the acquisition images methods were described by the computer-assistant tracking system. Fig. 1 shows the experiment system, whose main principle was digital image correlation technology. Then the problem of the parametrical characterization of soil microstructure was presented: geometry parameters and morphology parameters. To obtain these precise parameters would depend on the quality of images. On account of the concavo-convex feature of soil particles, the depth of field of the long work distance telecentric microscope lens had resulted in some blurred parts and clear ones existed in the same image. However, the greys of blurred parts of the image were changed smoothly. At the same time, the greys of clear ones were changed acutely. The greys of images were very like the frequencies of signals. So the wavelet transform was taken into account to improve the quality of such images. Based on the multi-resolution analysis of wavelet transform, the frequencies of signals were divided in different scales, In regard to the image transformed by the wavelet technology, the parts changed acutely of the image, that is the clear ones, were reserved. While, the parts changed smoothly of the image, that is the blurred ones, were abandoned. Therefore, the series images were acquired by auto-adjusting the control table of the computer-assistant tracking system under the same loading. Using the statistical characteristic of wavelet coefficient, the better quality images under the same condition were obtained by the fusion of series images. The experiment results show the method is valid and can improve the precision of the extraction parameters.