As a result of tracking a vehicle in the SD, we obtain the state-vector x as a function of a discretized time, namely a quintuple of the vehicle's estimated position coordinates in the road plane, its orientation , its speed v and its angular velocity around the normal to the road plane, i.e. the vertical axis of the vehicle, at each halfframe time point (i.e. every 20 msec). This Section illustrates the principle according to which motion verbs - conceptual descriptions - are associated with the geometric descriptions obtained by the tracking process. In order to simplify the subsequent explanation, we switch temporarily to a different image sequence, illustrated by the frame shown in Figure 25.
Figure 25: First frame of a video image sequence recorded from a high-rise building near the Ettlinger-Tor-Platz in Karlsruhe.
Figure 26 shows an enlargement of a region cropped from Figure 25 around the bus, with superimposed color-coded sections of the bus trajectory: each color corresponds to a particular motion verb referring to the road location occluded by the lamp post. We shall denote the activity described by a motion verb as an `occurrence'. The width of each color-coded trajectory section corresponds to a fuzzy membership function (a number between 0 and 1) characterizing the degree with which the algorithm associates the respective occurrence with the state vector at that halfframe time point. As long as this degree of certainty remains below a threshold for a particular motion, the trajectory will not be reproduced in the color corresponding to this occurrence. When this degree of certainty raises initially above the threshold, this will be indicated by open rectangles coded in the respective color as a tentative association between the trajectory section and the corresponding occurrence at that halfframe time point. If such a tentative association persists, it will after a while be converted into an accepted association with the degree of certainty shown in full color. In case the degree of certainty drops below a threshold, the visible association will be discontinued.
Figure: Enlargement of a region cropped from Figure 25 around the bus, with superimposed color-coded sections of the bus trajectory: each color corresponds to a particular motion verb referring to the road location occluded by the lamp post.
The occurrence `to drive across' nicely illustrates these three association phases: tentative, definitive with time-varying degree of certainty, and the gradual petering out once the bus has passed the location in question and the association `to drive across (the point on the road occluded by the lamp post)' no longer appears to be an appropriate one.
Agent Attributes Location Attributes Name Type A_speed
L_course L_distance in Pre_C M_C Post_C Pre_C M_C Post_C Pre_C M_C Post_C 1 Location Reference Occurrence: approaching
drive to location mut small -- small approaching -- approaching small decr small 0,2 2 Location Reference Occurrences: stopping
stop at location resul small decr null -- -- -- -- -- null 0,3
park at location perp null -- null -- -- -- null -- null 1,0
leave location inch null incr small -- -- -- null -- -- 0,3 3 Location Reference Occurrences: not stopping
arrive at location resul small -- small -- -- -- small decr null 0,1
run over location perp small -- small -- -- -- null -- null 0,1
depart from location inch small -- small -- -- -- null incr small 0,1 4 Location Reference Occurrence: passing
pass location term small -- small approaching ch receding small -- small 0,2 5 Location Reference Occurrence: receding
leave location behind mut small -- small receding -- receding small incr small 0,2