Recognition the Target Object Based on More Image Information

In robot visual perception system, the information for image segmentation is depth, color, shape, edge, motion and so on. Structured in an ideal environment, a single image feature can be separated from the target object and background, and, for example, in a single background condition, the target object color information can be extracted from the background. However, experiments in a complex environment, using single image information often can not complete the image segmentation; image information based on multi-object segmentation is the ideal way to solve the problem. This humanoid robot using stereo vision system, designed a fast image segmentation method gradually approaching the target areas that is to integrate depth, color and shape.