A method for detecting a person in a space includes at least one depth sensor producing spatial data about the space to be monitored, at least one sub-model, which is sub-dividable into further sub-models being used for at least one selected body part of a human, the spatial data being used to adapt the sub-models, the adaptation being checked by position parameters between the sub-models of different body parts, and the person being recognized using a complete model made up of the checked sub-models.