An automatic, quantitative image analysis system for construction materials

Dexiang Shi, Purdue University

Abstract

An image analysis software system has been developed for a UNIX operating system and a Tektronix terminal. It is designed for the particles and/or pores in construction materials. This system provides a means to measure many discrete fields of view, in order to obtain statistically meaningful results. The system can process images from either an electron or light microscope, or from an ordinary camera. Either negatives or positives can be used. A new method based on fuzzy probability has been developed to segment digital images into binary images. It is more consistent than existing techniques. The system uses object labeling. This makes spot filling, noise filtering, separation of touching objects, object counting and measurement of individual objects easier. The accuracy of the system has been verified with images having known geometric properties. Measurements have an error of less than 1% with an image of appropriate resolution. The system makes different measurements on separate particles and cut planes through massive samples. In the former case, the measurements are: the perimeters, areas, and maximum chords of particles in orthogonal directions, the maximum, minimum and mean values of these parameters, and their distributions. Also, the area of the image, the area faction of particles and the number of particles are measured. In the case of the cut planes, the estimated parameters are: volume fraction of particles, surface area per unit bulk volume, and surface area per unit particle volume. The size distribution, total number of particles and mean diameter can also be estimated. A new method based on computer simulation has been developed to estimate size distribution of particles of any modelable shape. Older methods have a tendency to give too large a number of small particles, and the new method does not.

Degree

Ph.D.

Advisors

Winslow, Purdue University.

Subject Area

Civil engineering

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server
.

Share

COinS