Disclosed herein are a system, method, and computer program product for updating a scene model (230) used for object detection in a video sequence by defining a relationship between a pair of mode models relating to different visual elements of said scene model (230). The method includes the steps of: determining whether the pair of mode models have a temporal correlation with each other, dependent upon a predetermined criterion (745); determining a classification of each mode model in the pair of mode models (740); modifying the relationship between the pair of mode models, dependent upon the determination of the temporal correlation and the determination of the classification (760); and updating the scene model based upon the modified relationship (770).