As soon as the association between geometric and conceptual descriptions at the level of occurrences has been stabilized, one realizes that this will not be sufficient: how could one describe more complex activities which might consist of a concation of many consecutive elementary occurrences?
Figure 32: Transition diagram showing the decomposition of the start symbol Drive_a_road_vehicle.
Figure 32 illustrates a sample decomposition of the complex activity `to drive a road vehicle' into a sequence of seven component activities. On the way to the highway it may often be advisable to fill the tank with gasoline. It thus appears appropriate to envisage that the (compound) activity `Drive_to_highway' contains a component activity `Get_gas' which in turn is detailed in Figure 33. In order to facilitate proper reference to the various locations in a gas station, Figure 35 sketches the layout of the gas station depicted in Figure 34.
Figure 33: Transition diagram - `Get_gas'.
Figure 34: Video frame of the gas station.
Figure 35: The layout of the gas station.
Figure 31 indicates occurrences which are relevant for driving on such a gas station premise. These occurrences have been grouped into a graph where links visualize potential transitions between occurrences. The lower right panel of Figure 31 shows the variation of the degree of certainty for these occurrences along the trajectory given by Figure 36 as a result of the tracking process described earlier. The degree to which the box in the graph shown in the top panel of this Figure around each occurrence is colored indicates the degree of certainty at the particular halfframe time point - in the particular case illustrated in Figure 36 it is the halfframe time point 1245 of this image sequence.
Figure 36: Halfframe time point 1245 of the gas station image sequence with overlaid trajectory and vehicle model.
Figure 37: The top part lists the halfframe time points for the beginning and ending of an acceptable association between a trajectory segment and an occurrence, followed in parentheses by the (system-internal) object identifier for the vehicle in question. The last column of this table indicates the level of specialization for this occurrence. The bottom panel contains the result of a process which converts the system internal representation of a concatenation of occurrences into a natural language text.
The concatenation of activities can now be converted into a coherent text as explained in (, ). The top part of Figure 37 lists the halfframe time points for the beginning and ending of an acceptable association between a trajectory segment and an occurrence, followed in parentheses by the (system-internal) object identifier for the vehicle in question. The last column of this table indicates the level of specialization for this occurrence as indicated by the graph shown in Figure 31. The text given in the bottom box of Figure 37 only uses the most specific occurrence in order to characterize the short compound activity of the vehicle on the virtual lane A.