The paper presents a system for target detection in static images based on the Hausdorff distance. The Hausdorff distance can determine the degree of resemblance between an image and a model. As such, the proposed system uses this distance as a cost function for the template matching task. The advantage of such approach is in the robustness, as the Hausdorff distance is quite tolerant of small perturbations in the image. Moreover, experiments with pedestrian detection shows that high detection rates can be achieved by using a generic human shape, thus avoiding sets that include large number of templates.