Abstract:

Objective Human body size data plays an important reference role in measuring human body type. Along with the continuous development of computer technology, non-contact anthropocentric methods based on two-dimensional images were reported to obtain human body size data, the accuracy of which depends largely on accuracy of the feature points of each part detected using these methods. For the shoulder, which is a feature part with fewer obvious feature point characteristics, the existing two-dimensional detection methods leads to large errors in the feature points, calling for further improvement.

Method In order to address the problems of false detection, missed detection, and large detection errors of the existing detection algorithms for shoulder feature points, an improved detection method for shoulder feature points based on human contour coding was proposed. The method was capable of determining the length of the feature chain code string by the size of the image occupied by the human body and selecting the feature points of the shoulder by a search method of feature partitioning, feature region traversal, and dynamic screening of the chain code string with geometric features. After acquiring the shoulder feature points, the shoulder width values were calculated based on the location of the feature points and compared with the actual manually measured shoulder width values to analyze the experimental results.

Results In this research, a total of 100 young testers of different body types (50 males/50 females) were selected to test the feasibility of the method under the same shooting environment. The experimental study showed that the method can quickly and accurately identify the shoulder feature points of the human body, and the locations of the selected feature points exist and are single, reducing the occurrence of false detection and omission (Tab. 1). The average error derived was only about 3%, which verifies the feasibility of the method.

Conclusion While maintaining the advantage of the low cost of the 2-D measurement method, it also greatly improves the efficiency of 2-D human shoulder width size detection, enabling it to provide more accurate data support in the subsequent human body shape research, clothing matching and professionalization, and other related fields.

Key words: shoulder feature point, acromial point, human contour coding, body size measurement, feature chain code string, shoulder width