3D defect segmentation on CT volume data using morphology and resample techniques.

Area of Science:

• Image processing
• Non-destructive testing
• Materials science

Background:

• Accurate segmentation of Computed Tomography (CT) volume data is crucial for non-destructive testing and evaluation.
• Existing methods can be influenced by artifacts, leading to inaccurate defect identification.

Purpose of the Study:

• To propose and validate a novel two-step approach for 3D defect segmentation in CT data.
• To effectively eliminate segmentation artifacts and false positives.

Main Methods:

• An initial segmentation is performed using a 3D morphological method.
• A resampling in polar coordinates method is applied to refine the segmentation.
• The approach is tested on CT volume data, including noisy datasets.

Main Results:

• The proposed two-step method successfully segments 3D defects.
• False segmentation and artifacts are effectively eliminated.
• The method demonstrates robustness and usefulness even with noisy CT data.

Conclusions:

• The novel two-step segmentation approach is effective for accurate 3D defect identification in CT data.
• This method enhances the reliability of non-destructive testing and evaluation by removing false positives.
• The technique shows promise for applications involving noisy CT volume data.