| Abstract: | A model-based approach to determining the location of an automated guided vehicle (AGV) in the navigation session is proposed. Significant scenes along the navigation path are taken and stored as model images. Also, specific templates of each model image are chosen in advance for matching. When the AGV approaches one significant scene in the navigation session, an input image is taken and the corresponding model image is fetched. Each template in the model image is matched with the input image through the use of the three-step method. The templates are directly used for matching; therefore, no special image processing techniques are required for the input image in the navigation session. In addition, several templates in a model image are used for matching instead of the model image itself. This promotes the robustness of the matching results and saves the computation cost. Furthermore, the 2D string method is used as a post-verifier to discard erroneous matching results. The remaining correct matching results then can be used to determine the vehicle location on the navigation path. Simulated experiments are implemented and the results confirm the feasibility of the proposed approach. © 1994 John Wiley & Sons, Inc. |
