Uncertainty reasoning in hierarchical visual evidence space
暂无分享,去创建一个
One of the major problems in computer vision involves dealing with uncertain information. Occlusion, dissimilar views, insufficient illumination, insufficient resolution, and degradation give rise to imprecise data. At the same time, incomplete or local knowledge of the scene gives rise to imprecise interpretation rules. Uncertainty arises at different processing levels of computer vision either because of the imprecise data or because of the imprecise interpretation rules. It is natural to build computer vision systems that incorporate uncertainty reasoning. The Dempster-Shafer (D-S) theory of evidence is appealing for coping with uncertainty hierarchically. However, very little work has been done to apply D-S theory to practical vision systems because some important problems are yet to be resolved.
In this dissertation, a computational framework is presented to show how Dempster-Shafer evidence theory can be applied to a hierarchically structured hypothesis space in a computer vision application. Based upon a priori knowledge, uncertain visual information is transformed from one level to another through stages of evidence collection and mapping, hypothesis generation, evidence interaction, and hypothesis verification. The system reasons about significant perceptual features through both top-down and bottom-up active processes. It is shown how to make use of partial and locally ambiguous information at different levels of abstraction to achieve reliable interpretation. It is also shown how the reasoning process can make use of spatial relationships among pieces of visual evidence to strengthen the reasoning results. Some theoretical problems, which arise from the adoption of the Dempster-Shafer model as the paradigm for a computer vision system, have been resolved. Methods for implementing this algorithm are presented for applications in object recognition and image understanding. Experiments on some applications are given to demonstrate the merit of the framework.