Abstract: The system developed in this paper works in two phases.In the learning phasea scene containing a single object is shown once at a time.The range images of ascene are obtained by rangcfinder.The description of each scene is built in terms of properties of regions and relations between them.This description is stored as an object model.In the recognition phase,an unknown scene is described in the same wayas in the learning phase.Then the description is matched to the object models so that stacked objects ale recognized sequentially.Efficient matching is achieved by a com-bination of data-driven and model-driven search processes.Experimental results has shown that this approch is promissing.