Automated positioning of breast feature points for parameter extraction based on 3D point cloud

• Published in: International journal of industrial ergonomics Vol. 107; p. 103759
• Main Authors: Zhong, Zejun, Zhang, Beibei, Gu, Bingfei, Sun, Yue
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.05.2025
• Subjects: Automatic positioning; Breast shape; Feature points; Morphological parameters; Predicted models; Breast shape; Automatic positioning; Morphological parameters; Predicted models; Feature points
• DOI: 10.1016/j.ergon.2025.103759
• ISSN: 0169-8141

To achieve automatic extraction of parameter for female breast shape analysis, this paper proposed a “point-parameter-type” method based on 3D point cloud data. To standardize the measurement method, nine feature points (i.e., BBP, BP, FAP, FNP, LUBP, LBP, MBP, RUBP and UBP) and three lines (i.e., BBL, BL and BC) were firstly defined according to the characteristics of breast shape. Utilizing the 3D point cloud data, four positioning methods, including Max-Distance, Inflection-Points, Slope and Intersection-Point, were proposed to automate the positioning of feature points. Finally, breast morphological parameters for shape classification were calculated or predicted using computational models, and 140 subjects were randomly selected to verify the method accuracy. The results indicated that the recognition accuracy rates were 94.74 % for type FC, 89.06 % for type UO, and 89.47 % for type PO, demonstrating that this method is feasible. This study aims to establish a foundation for the automatic measurement of breast and provide valuable support for bra size recommendation for consumers during online shopping. •We proposed a “point-parameter-type” method for automatic breast shape analysis.•Feature points and lines were standardized based on the definition of female breast.•Four methods were proposed to position the feature points based on 3D body data.•Our method achieved around 85.71 % accuracy rate in breast morphological recognition.•Error analysis in extracted and manually measurements showed the feasible of methods.