The Object Tracking System Of Robot With Omni-directional Vision

The object tracking system by vision of mobile robot can work well in or out room, not to use auxiliary facilities. The lens with small angle of view used before inclined to miss objects. If observering objects in a big area, we must use a stent with many freedoms. The object tracking system with omni-directional vision can capture information of hemisphere field. This is important for vision tracking.Recently many object tracking methods can be applied on Omin-Vision Mobile Robots Platform which performs differently for specific condition. The methods based on a single kind of characteristic information usually fail under a complex condition. This article describes an object tracking compound method combining much information of the object. With a model matching algorithms, a snake model algorithms based on contour information and an algorithms based on distinguished points, the particle filter algorithms is at center position. At the same time all the methods are evolves to a certain extent A distributed and parallel architecture is finer for the compound method. The particle filter algorithms acts the role of data fusion center which produce an output that indicate the area where other three methods detect the object. The three non-center methods produce an output respectively that gives a feedback to the particle filter method. It is easy to develop parallel program and to exploit multi-trackor on mobile robots platform for the architecture. We can also make only one or two methods valid in specific environment. The compound method overcomes the shortcoming of unprecise locating and big randomicity for the particle filter method and complex calculation and bad real-time for other three methods. The result of tracking many objects suggests that the compound method which unites the virtue of other methods have a good robusticity and accuracy.