Identification of concrete aggregates using K-means clustering and level set method

Digital image processing techniques such as crack detection, disaster fasting and aggregate recognition have been widely used in civil engineering. However, how to robustly recognize low-contrast images is still a changing work. In this work, recognition of concrete aggregates of SEM or Microscope images with low contrast is implemented by the K-means clustering and level set method (LSM). The results show that K-means clustering can be used in the recognition of RGB images or gray images with different gray levels while LSM can be used to recognize aggregates in lowcontrast images with a simple gray level. Comparatively, LSM exhibits a higher accuracy than K-means clustering. For low contrast images with different gray levels, a combination of K-means clustering and LSM is more efficient for aggregate recognition in images when compared to other methods.

Automated summary

In this work, recognition of concrete aggregates of SEM or Microscope images with low contrast is implemented by the K-means clustering and level set method (LSM). The results show that LSM exhibits a higher accuracy than K-means clustering, and a combination of K-means clustering and LSM is more efficient for aggregate recognition in low contrast images.