一种基于图像匹配的闭环检测方法

An approach to loop-closing based on images matching

为提高较大规模环境下同时定位与地图创建闭环检测的鲁棒性和实时性,提出一种基于单目视觉的闭环检测方法。利用尺度显著算法选择图像的感兴趣区域,对显著特征区域再选择后,通过提取图像尺度不变特征进行图像匹配以实现闭环检测,解决了特征匹配计算量过大对数据关联的速度影响问题。

To improve the robustness and real time of loop-closing of simultaneous localization and mapping in larger scale envi- ronment, a mono-vision based approach is proposed. The visually salient regions are selected from an image using scale saliency. algorithm. After the most salient regions are re-selected, the scale invariant feature transformation features extracted from the salient regions in current image are used to match against the possible candidate features, and thus the problem of high computation cost of features matching is well resolved.