Targets image segmentation based on adaptive Mean-Shift algorithm for surface moving platform

The visual system of surface moving platform, such as exploration ship, military vessels, unmanned boat etc., is an important image information acquisition equipment for surface targets, meanwhile, image segmentation is an important step to separate targets from background of image for target feature extraction. Because the targets included in visual frames collected from surface moving platform visual system are different with the difference of distance and size, the classical Mean-Shift image segmentation method based on a global bandwidth needs to change the values of bandwidth constantly, which needs to adjust parameters in handwork when the contrast ratio of target and background is changing. So a kind of adaptive image segmentation algorithm based on improved Mean-Shift is proposed. Firstly, spatial bandwidth is adaptively computed according to the estimation of gray distribution around the reference point. And then the gray-level bandwidth is adaptively computed with a novel Bayesian theory in the corresponding windows. In the simulation experiment, the target frames like both close and distant scenarios or different contrast scenarios, which extracted respectively from the video sequence are used. Experiment results shows that proposed segmentation algorithm has high accuracy and is suitable for the segmentation of surface targets with different distance.