TY - CONF A1 - Siefke, Lennart A1 - Linden, Anna T1 - Towards a robust automated surface inspection method for CT-scanned cannulas N2 - For certain cardiovascular diseases, cannulas are implanted into the blood circuit. To match the patients individual anatomy of the heart, there is research for cannulas to be custom-designed and manufactured aided by 3D printing. However, cannulas have to hold very high standards with regard to the smoothness of their surfaces, as rough patches can lead to formation of blood clots. Therefore, this work uses computer vision to detect such patches as part of quality assurance. First, the produced cannula is scanned using a precise CT scanner and transformed into a 3D mesh object. Rough patches in an otherwise smooth but curved surface are detected by using cosine similarity between neighboring faces and a statistical evaluation. In the end, this method is able to raise a warning when curved surfaces are not smooth enough and visualizes the problematic patches. However, there is just limited access to test data currently and the scanner used needs to be upgraded. T2 - 3D in Science & Applications (3D-iSA) 2024 CY - Berlin, Germany DA - 26.11.2024 KW - Algorithm KW - Additive manufacturing KW - Surface evaluation PY - 2025 UR - https://www.gfai.de SN - 978-3-942709-34-7 SP - 66 EP - 70 AN - OPUS4-63057 LA - eng AD - Bundesanstalt fuer Materialforschung und -pruefung (BAM), Berlin, Germany ER -