Employing flowgraphs for forward route reconstruction in video surveillance system

摘要

Pawlak’s flowgraphs were utilized as a base idea and knowledge container for prediction and decision making algorithms applied to experimental video surveillance system. The system is used for tracking people inside buildings in order to obtain information about their appearance and movement. The fields of view of the cameras did not overlap. Therefore, when an object was moving through unsupervised areas, prediction was needed to identify the same object in the adjacent camera. Moreover, while being unobserved, the object was represented by measurements of probability of appearing in the subsequent camera’s field of view. Those probability values were obtained on the basis of knowledge from the past events contained in the flowgraph. Backward Route Reconstruction methods generated paths of objects providing input for prediction of next steps in the path. This prediction was named as Forward Route Reconstruction. The output of the prediction algorithm is a tree of probabilities of future object movement. Methods for creating a flowgraph from the paths of objects and for obtaining probability values for movement prediction are presented in this paper together with some experimental results and discussion.