Perceptual Organization and Visual Recognition

摘要: A computational model is presented for the visual recognition of three-dimensional objects based upon their spatial correspondence with two-dimensional features in an image. number components this are developed further detail and implemented as computer algorithms. At highest level, a verification process has been which can determine exact values viewpoint object parameters from hypothesized matches between image features. This provides reliable quantitative procedure evaluating correctness interpretation, even presence noise or occlusion. Given method final evaluation correspondence, remaining system aimed at reducing size search space must be covered. Unlike many previous approaches, does not assume that it possible to directly derive depth information Instead, primary descriptive component perceptual organization, relations detected among basic requirement organization should accurately distinguish meaningful groupings those arise by accident position. used constraints satisfied algorithms grouping. specific algorithm problem segmenting curves into natural descriptions. Methods also using viewpoint-invariance properties infer The itself described, both covering range viewpoints objects. evidential reasoning combine multiple sources most efficient ordering search. use allows automatically improve its performance gains experience. In summary, shown practical basis current systems provide promising path development improved capabilities.