Primary al grain size measurement of AlSi hypo-Eutectic alloy using mathematical morphology algorithms

Yan Xu, Naoya Hirata, Koichi Anzai

Research output: Contribution to journalArticlepeer-review


Homogeneous distribution of primary Al phase grain (hereafter, primary Al grain) size in an AlSi alloy casting is desirable. Traditional image processing techniques found difficulties in making proper image segmentation for the individual primary Al grains. Recently, a new image processing technique, mathematical morphology, is paid attention to because of its capability for a flexible image processing. In this paper, an image processing method based on mathematical morphology algorithms was proposed for the appropriate image segmentation and measurement of primary Al grain size in aluminum alloys. Opening algorithm was applied to identify primary Al grains and Watershed transformation (WST) using Euclidean distance map (EDM) as marker-image was used to separate individual primary Al grains. Finally, a second implementation of WST was used to classify the isolated small primary Al grains that had not been individually identified in the first time WST. The results showed that the proposed method appropriately identified primary Al grains with enough quality to evaluate the size distribution. The quality of primary Al grain size measurement result was as high as human operations. Meanwhile, the proposed method also was more efficient than the subjective measurement by hands.

Original languageEnglish
Pages (from-to)1288-1295
Number of pages8
JournalMaterials Transactions
Issue number8
Publication statusPublished - 2018


  • AlSi hypoeutectic alloy
  • Image segmentation
  • Mathematical morphology
  • Primary Al grain size measurement
  • Semi-solid slurry

ASJC Scopus subject areas

  • Materials Science(all)
  • Condensed Matter Physics
  • Mechanics of Materials
  • Mechanical Engineering


Dive into the research topics of 'Primary al grain size measurement of AlSi hypo-Eutectic alloy using mathematical morphology algorithms'. Together they form a unique fingerprint.

Cite this