Omnidirectional vision (Omni-vision) has the feature that an extremely wide view can be achieved simultaneously.
The omni-image brings a highly unavoidable inherent distortion while it provides hemispherical field of views. In this
paper, a method called Spherical Perspective Projection is used for correction of such distorted image. Omni-vision
target recognition and tracking with fisheye lens for AGVs appears definite significant since its advantage of acquiring
all vision information of the three-dimensional space once. A novel Beacon Model and Omni-vision tracker for mobile
robots is described. At present, the research of target model has many different problems, such as outdoor illumination,
target veiling, target losing. Specially, outdoor illumination and beacon veiling are the key problems which need an
effective method to solve. The new beacon model which features particular topology shape can be recognized in the
outdoors with part veiled of the object. In this paper an improved omni-vision object tracking method based on mean
shift algorithm is proposed. The mean shift algorithm which is a powerful technique for tracking objects in image
sequences with complex background has been proved to be successful for the fast computation and effective tracking
problems. The recognition and tracking functions have been demonstrated on experimental platform.