Digital image matching plays an increasingly important role in digital photogrammetry. In low-level vision, the method of image matching with simultaneous object reconstruction by least squares matching is the most common and comprehensive approach to the problem. In the matching process object points are represented by small surface patches, thus the results of object based image matching procedures are coordinates of object points or groups of points describing the object surface. This kind of object description is appropriate with topographical applications, as a digital elevation model can be derived in the matching process. The representation of industrial objects by sets of points is, however, not suitable for visualization or easy interpretation. It has to be transformed into a more convenient description, like a CAD surface description, in a secondary step. In the scope of this paper, an object based matching algorithm is presented, where the object surface as a whole is matched in image and object space. Methods are introduced to reduce storage requirements and calculation times for the setup and solution of the normal equation system.