Archival and Retrieval of Traffic Video Shots: an Approach Using TTSR

In this paper, a model for archival and retrieval of traffic video shots is proposed. It is proposed to represent the content of traffic shots by higher level features instead of low-level features. The semantic information of the frame sequence of each shot are studied. The semantic information are represented by perceiving the topological-triangular spatial relationships (TTSR) existing among moving vehicles with additional parameters such as total number of moving vehicles and total number of each category of moving vehicles. The represented traffic shots are archived in the knowledge base. Given a query traffic shot, a matching strategy is designed which supports effective retrieval of similar shots for a given query. The traffic shots are ranked and retrieved based on the proximity values.

Keywords:

semantic information , topological-triangular spatial relationships (TTSR) , matching , archival , retrieval , traffic video shots