Mechatronic Design and Experimental Research of an Automated Photogrammetry-Based Human Body Scanner

Abstract

The article concerns the mechatronic design and experimental investigations of the HUBO automated human body scanning system. Functional problems that should be solved by using the developed scanning system are defined. These include reducing the number of sensors used while eliminating the need to rotate a human and ensuring the automation of the scanning process. Research problems that should be the subject of experimental research are defined. The current state of the art in the field of systems and techniques for scanning the human figure is described. The functional and technical assumptions for the HUBO scanning system are formulated. The mechanical design of the scanner, the hardware and information system architectures as well as the user’s mobile application are presented. The method of operation of the scanning system and its innovative features are discussed. It is demonstrated that the developed solution of the scanning system allows the identified problems to be overcome. The methodology of the experimental research of the scanning system based on the photogrammetry technique is described. The results of laboratory studies with the use of dummies and experimental research with human participation are presented. The scope of the research carried out allows answers to the identified research problems related to the scanning of the human figure using the photogrammetry technique to be obtained. As part of laboratory tests using a measuring dummy, a mean error of 0.65 mm and standard deviation of the mean of 0.65 mm were obtained for the best scanner configuration. Research with human participation was carried out for the scanner version, in which the scanning time was 30 s, with the possibility of its reduction to 15 s. The results of studies using realistic dummies and with human participation were compared using the root mean square error parameter (RMSE) provided by the AliceVision framework, which was available for all analyzed objects. As a result, it was observed that these results are comparable, i.e., the RMSE parameter is equal to about 1 px.