This paper proposes a new method of geometry modeling for robotic applications. We utilized visual ID-tags to acquire the tangent planes of objects, and generated polyhedral approximate models of the environment. We developed the optimization algorithm for estimating the tag configuration and the method for forming polyhedral models of objects. Our approach enables an easy modeling using a single camera without pose information. The geometry models can be superimposed onto the real environment, and robots use the models in their tasks. We examined the performance of our modeling method, and demonstrated the availability of the models to robotic applications.