Hierarchical Method for Stereophotogrammetric Multi-object-Position Measurement

The classical stereophotogrammetric methods based on area correlation are relatively slow if the whole image is analyzed. The new proposed method differs from classical stereophotogrammetric methods in that a hierarchical structure is incorporated in the procedure, so that real-time processing is possible and the relative error is kept reasonably constant even with large variations in one direction (e.g. in road traffic analysis). This is achieved by adapting image resolution to distance. Computation costs are significantly reduced. The method is very suited for implementation in hardware; it runs in real time and can be applied to moving objects that are automatically segmented. The aim of this research project is to reduce the computation power needed although a complex quality criterion is used.