Tracking Method Research Of Multi-actors Based On Dynamic Bayesian Network

The analysis of the human's motion is a hot topic in the field of computer visionbecause of its'application. The researchers in this field mainly focus on the detection,tracking of moving actors and the identification of simple acts of human. The key pointis how to detect and track the actors accurately and effectively in the video imagesequence.In the monocular video there are tracking difficulty of multiple view angles aboutmulti-actors due to the impact of the shelter and the deepness. In this paper, a method isproposed firstly to distinguish the different view angles of moving actors in the samescene. And at the same time, some study shows the actors'figure features can trackmoving actors accurately within one view angle. So the arithmetic mentioned in thispaper put forward a united model based on DBN (Dynamic Bayesian Network),according to the combination of the actors'figure features and the spatial motorialfeatures, which aim to solve the tracking problem of multi-actors whit multiple viewangles in one same scene.One analysis system of human's motion was introduced to validate the arithmeticmentioned in this paper. The results of the experiments showed that this approach haveobtained satisfied effect in the adaptability and the veracity. This approach has somereference value to realize the tracking success of multi-actors.In the step of the Feature Matching, some constraint conditions are used to reducethe effect of the small differentia. How to refine such characteristic parameters of thealgorithm and improve the approach's adaptation and facility is the important directionin the future research.