To facilitate the design of the serialized implants to satisfy the requirements of the population, a novel method is put forward for constructing an average bone model (ABM) with semantic parameters as template. Firstly, ABM is created from the existing bone models, among which each bone has an equal contribution to the ABM. Secondly, combined with medical semantics, some characteristic points and semantic parameters are defined on ABM, and then parameter values for each bone can be automatically obtained through its registration and deformation to the ABM. Lastly, an average bone template (ABT) is constructed by configuring the semantic parameters and building the constraints between parameters. Taking 100 femur models as samples, we construct an average bone template, and the template can be easily extended to generate a new average template.