This figure shows the 400 silhouette prototypes and the corresponding 400x400 co-occurrence matrix. The resulting classifier classifies tracked objects as individuals, groups of people, clutter/lighting effects, and vehicles. There are relatively few false positives. The histograms on the right show the activity over a single day. Morning and evening rush hours are evident for pedestrians, groups, and vehicles. A midday rush hour is prominenant for both individuals and groups of pedestrians.
Below are blown up versions of the images of example images from each classified track.
