Detecting the dislocations in metal crystals from microscopic images

摘要

Studying the distribution of dislocations in metal crystalline structures has been an important subject in metallurgic science for many years. In this paper we describe a computer vision based process which automatically extracts dislocations from one type of microscopic images which contain such dislocations. In these microscopic images, dislocations appear as dark and triangularly shaped regions. However, some of them are lumped together and form regions of complex shapes. In the suggested process we first extract the boundaries of all the dark regions. Next we trace along the extracted boundaries to detect possible corners and edges of the dislocations. Lastly, we determine the existence of the dislocations and their locations by detecting properly spaced and oriented corners and edges. Our experiments have shown that this process is very effective. The approach is comprehensive and deals with many important issues in images processing and pattern recognition.