摘要 |
The invention relates to a sequence of depth images (B') showing at least one person (4) from the front. At least one region (9) of person-similar contour is determined in the depth images (B') and allocated to the person (4). A basis point (10, 10') of the region (9) of person-similar contour is determined in space and an offset vector (11, 11') is determined. The offset vector (11, 11') proceeding from the basis point (10, 10') has a component pointing forward in reference to the person (4). A boundary (14) of the space separates a part of the region (9) of person-similar contour (common arm region) from the basis point (10, 10'). The border (14) comprises the offset vector (11, 11') and runs mirror-symmetrically to a vertical plane comprising the offset vector (11, 11'). Each point of the border (14) is spaced apart from the basis point (10, 10') at least by the length of the offset vector (11, 11'). Individual arm regions (16, 17) are determined within the common arm region. Said regions are determined by those points (18) in three-dimensional space which have a distance smaller than a minimum distance to at least one other point (19) of the same individual arm region (16, 17) and a distance greater than a maximum distance to all other points (20) of the common arm region. The minimum distance is approximately the same as a grid dimension of the space. The maximum distance is greater than the grid dimension. Further evaluations of the depth images (B') are made based on the individual arm regions (16, 17). |
申请人 |
SIEMENS AKTIENGESELLSCHAFT;FIEGERT, MICHAEL;FRIESE, THOMAS;VON WICHERT, GEORG |
发明人 |
FIEGERT, MICHAEL;FRIESE, THOMAS;VON WICHERT, GEORG |