Bayesian analysis for fusion of data from disparate imaging systems for surveillance

摘要

This paper investigates whether the segmentation of surveillance images can be improved by fusing low-spatial-resolution thermal data with high-spatial-resolution visual information. The context of this investigation is the surveillance of sterile zones where an alarm is required should a person enter the zone and at no other time. The aim is to reduce false alarms due to wildlife movement or changes in environmental conditions. A calibration algorithm has been developed which maps correspondence between the two cameras. By concentrating the area of search on that indicated by the thermal camera, the analysis in both images can then be made more elaborate without undue computational effort. Segmentation of the highlighted object is achieved using Markov Random Fields.