中文摘要：

针对蛇形搜救机器人在复杂颠簸的搜救场景中连续获取的单帧图像之间旋转变化剧烈的特点,提出一种结合ORB（oriented FAST and rotated BRIEF）特征检测算子和局部敏感哈希（locality-sensitive hashing,LSH）特征关联算法来完成蛇形搜救机器人的同步定位与建图（simultaneous localization and mapping,SLAM）。该方法具有尺度不变性和旋转不变性,可有效解决特征点的检测与匹配问题。实验平台采用自主研制的具有高清摄像头的蛇形搜救机器人,分别对不同步态、不同场景进行实验验证,结果表明,与传统视觉SLAM相比,该算法计算量小,时效性强,适用于复杂环境下蛇形搜救机器人的工作。

英文摘要：

As the image search and rescue scene has acquired between neighboring frames by characteristics of high rota ciation algorithm is adopted to complete simultaneous 1 tional changes, ocalization and snake-like rescue robot in complex bumpy ORB feature detection and LSH feature asso- mapping of the snake-like rescue robot. This method is invariant to scale-invariant and rotation-invariance, it can salve the feature point detection and matching problem effectively. Experimental platform used independently developed snake-like rescue robot with a high defini- tion camera, experiment with different scenarios and different movement types, the results demonstrate that, com- pared with the tradition visual SLAM, the algorithm is of small amount of calculation and high real-time, it is suit- able for snake-like rescue robot applications in search and rescue environment.