Automated Porosity Segmentation

Dragonfly provides automated thresholding methods to reliably segment porosity in single-material samples, such as castings, with closed pores. This lets you quickly and accurately prepare data for analyzing pores with parameters such as volume fraction percentage, pore size distribution, sphericity, and others.

Right-click the image data that you need to segment and then choose Automated Porosity Segmentation in the pop-up menu. You can then choose a method — Basic Otsu, Advanced Otsu, or Sobel Threshold — to automatically segment the pores in the sample. The result of an automated porosity segmentation is shown below.

Automated porosity segmentation (colored red) of an automotive part

The following options are available for applying automated porosity segmentation. In all cases, you should use the method that works best for your sample type and scan quality.

Basic Otsu… This is the fastest available method and typically captures the largest pores in the sample but may not detect smaller pores. The Basic Otsu method includes creating a region of interest for the air and void space based on the lower Otsu threshold for the full 3D dataset. This ROI is then refined to remove segmented exterior air using the 'Process Islands' tool to remove the largest component within the ROI.

Segmentation of pores according to the Basic Otsu method for a casting dataset

Advanced Otsu… This fast and reliable method typically captures smaller pores than the Basic Otsu method and creates two regions of interest that describe the pores and material in the sample respectively.

In this method, a region of interest is first defined for the material using the upper Otsu threshold. Internal pores are then filled and a dilate function is applied with a 3D kernel of size 5 to increase the ROI size beyond the edge of the segmented material. The ROI is then split at the Otsu threshold to effectively separate the pores from the segmented material. The layer around the material in the resulting 'Porosity ROI' is then removed with the 'Process Islands' tool.

Segmentation of pores according to the Advanced Otsu method for a casting dataset

Sobel Threshold… This advanced method is capable of finding the smallest pores in a sample and can be applied to challenging data. For example, in cases in which brightness variations are present across the image and in cases of low signal to noise. For the segmentation step, this method uses Sobel image filtering without any pre-smoothing to highlight the edges of pores. This initial ROI is then 'closed' and inner areas are subsequently 'filled', while the edge that defines the material is removed.

Segmentation of pores according to Sobel method for a casting dataset shown