Indoor robot vision positioning method

The invention relates to an indoor robot vision positioning method and is applicable to robot navigation. By simulating human vision identification process modeling, a robot is subjected to learning training and is enabled to achieve positioning independently so as to identify own positions, and the algorithms such as a visual saliency model and a gist model are integrated to form a complete set of vision positioning algorithms. Compared with an existing vision positioning technology, the indoor robot vision positioning method has the advantages of higher accuracy rate and shorter computing time and training time. 本发明涉及一种机器人室内视觉定位方法，适用于机器人导航。本发明通过模拟人类视觉识别过程建模，对机器人进行进行学习训练，使之可以自主实现定位，识别其自身所处位置，采取视觉显著性模型(Saliency Model)、场景主旨模型(Gist Model)等算法，结合集成为一整套完整的视觉定位算法，与现有的视觉定位技术相比，具有更高的准确率，同时减少了运算时间和训练时间。