| 摘要 |
A method for comparing geometric shapes to each other is disclosed. When used for vehicle positioning, the present method determines which map path of several candidate map paths best matches an actual path traveled by the vehicle as measured by sensors. The vehicle path and each candidate map path are generalized to a given degree of generalization, thereby yielding an overall trend of the vehicle path and overall trends of each of the candidate map paths. The trend of the vehicle path is compared to the trend of each of the candidate map paths. Based on these comparisons, one or more candidate map paths may be eliminated. If more than one map path remains, the vehicle path and each of the remaining map paths are generalized again, this time to a lesser degree of generalization, and comparisons are made between the trend of the vehicle path and the trend of each of the remaining map paths. Based on these comparisons, one or more map paths may be eliminated. These steps are repeated until only a single map path remains. The vehicle is determined to be located on the road represented by the remaining map path. The multi-resolution trend metric concept can be applied to valuation of the accuracy of a geographic database. A generalized version of a sampled geographic feature is compared to one or more generalized versions of ground truth of the feature to determine how well the generalized versions match and how the shape error between the sampled and ground truth segments varies as a function of resolution.
|
