The alignment method refers to the idea of superposing a model onto an image. The problem is separated in two parts, i) first find a possible transformation of the model on the image ii) then validate this transformation by aligning the model onto the image.

We propose an extension of the traditional alignment method in order to deal with complex objects sustained by projective mappings. Much of the work is focalized on how to cope with spurious data as long as processing time is not becoming too excessive. Moreover, the robustness and simplicity of the approach make it appealing for a broader class of objects, not necessarily urban parts.