Recovering Non-overlapping Network Topology Using Far-field Vehicle Tracking Data

This paper presents a weighted statistical method to learn the environment´s topology using a large amount of far field vehicle tracking data collected by multiple, stationary non-overlapping cameras. First, an appearance model is constructed by the combination of normalized color and overall model size to measure the moving object´s appearance similarity across the non-overlapping views. Then based on the similarity in appearance, weighted votes are used to learn the temporally correlating information and hence to estimate the mutual information. By exploiting the statistical spatio-temporal information, our method can automatically learn the possible links between disjoint views and recover the topology of the network. The effectiveness of the proposed method is demonstrated by experimental results both on simulated and real video surveillance data