| Video surveillance system, as an important technique of advancedvideo technology, is of increase importance recent years. It is alwaysdesirable to develop friendly, effective and intelligent video surveillancesystem for handling various scenarios. In this paper, we focus on two keytechnologies of video surveillance system. One is video enhancement andthe other is behavior recognition in intelligent video surveillance system.We first proposed a new video enhancement algorithm. The proposedalgorithm combines new inter-frame-constraints (ECB) and anintra-frame-constraints-based (ACB) enhancement algorithm to achieveappealing view of video sequences. In one hand, the ACB enhancementalgorithm adjusts the color tone according to the features of some interestregion to make a frame look visually appealing. On the other hand, the interframe constraints can make different frames look more consistent. Theproposed algorithm can achieve appealing visual experience for users.As to behavior recognition issue, we proposed an extendednetwork-transmission-based (ENTB) algorithm. The network modelconsiders the entire scene as an error-free network. In this network, eachnode corresponds to a patch of the scene and each edge represents theactivity correlation between the corresponding patches. Based on thisnetwork, we further model targets in the scene as packages while activitiescan be modeled as the process of package transmission in the network. OurENTB can deal with arbitrary segmentation methods. We further proposedan iterative training method to calculate the transmission energies betweenpatches. The proposed ENTB algorithm can be used for complex group activity. By applying the ENTB model, we can effectively recognizescene-related abnormal activities as well as target-related group activitiessuch as approach, leave, fight and etc. |
PDF Full Text Request