A crucial part of the aerotriangulation process is the identification and precise measurement of well-distributed points over areas with multiple image overlap. The selected point positions should be determined with sub-pixel accuracy on all the overlapping images. In this paper we describe an algorithm which combines gray-scale matching with geometrical constraints. The least squares method works in object space which is derived from the overlap and sensor geometry approximations. The model is based on the simultaneous use of multiple images which is fundamentally different from the traditional pairwise approach. Some practical data highlight the object-space-constrained, affine-transformation-based, gray-scale, least squares matching method.