未知环境中基于视觉显著区域的拓扑定位

Topological localization based on visual salient regions in unknown environments

针对移动机器人在未知环境中的导航问题,提出并实现一个新的基于视觉显著区域的拓扑定位系统。首先采用中心—周围差方法在多尺度图像空间中计算颜色及纹理对比,根据检测出的显著线索构造适宜尺寸的显著区域。然后将这些场景中的视觉显著区域利用隐马尔科夫模型组织成为拓扑图中的一个顶点,从而将定位问题转化为隐马尔科夫模型(HMM)的估值问题。该系统支持机器人在线建立环境的拓扑模型,同时进行定位。实验结果表明,该方法能够在机器人移动过程中发生尺度、2维旋转、视角等变化时稳定地检测出显著视觉区域,场景识别率较高。实验证明该定位系统有能力保证机器人在未知环境中的安全导航。

This paper presented a new topological localization system for mobile robot navigation based on salient visual regions. These salient regions were obtained by computing the contrast of color and texture among multi-scale image spaces, and then they were organized to construct the vertex of topological map by using HMM（Hidden Markov Model）. So localization problem can be converted to the evaluation problem of HMM. In our system, the topological map of environment can be created online and the robot locates itself concurrently. Experiments show that the salient regions are stable to large changes in scale, 2D rotation and viewpoint. Higher ratio of recognition was obtained. And our localization system can guarantee mobile robot navigation safely in unknown environment.