针对现有导盲设备存在的检测效果差, 系统复杂, 不便携带等问题, 开发了基于单一Kinect传感器的导盲系统。该系统能够同时检测地面和头部之间的所有障碍物, 并将准确的路况信息通过语音传递给用户, 以实现导盲作用。该系统运用深度图像翻转算法和障碍物识别算法来识别地面障碍物、地面坑洞、悬空障碍物3种类型的障碍物; 同时运用深度图像过滤算法和障碍物距离计算算法来确定有效范围内障碍物的位置。设计制作了导盲试验样机, 在复杂室内环境多个场景下对障碍物的识别和检测进行了实验测定。实验结果表明: 该导盲系统能够在一定范围内较为准确地识别面积大于10-3 m2的障碍物及高度差超过0.02 m的坑洞, 经过图像过滤可将用户身高以内, 宽度1 m内的障碍物的最佳避障路线通过语音传递给用户, 满足了导盲需求。该系统的应用不仅对导盲具有重要作用, 对机器人自主行走也有参考意义。

A blind navigation aid system was designed to improve the poor detection, complex structure and inconvenient to carry for the existing guide equipment. The system could detect all the obstacles between the ground and the user's head, and could deliver accurately the traffic information through a voice prompt to the user, so as to complete the navigation. The depth image rotation algorithm and detection obstacle technology were applied in this system to identify three types of obstacles including ground obstacles, ground pits and hanging obstacles. Meanwhile, the depth image filtering algorithm and obstacle distance calculation were used to determine the position of the obstacle in an effective range. A prototype was developed, and the experiments on the obstacle recognition and obstacle position detection were carried out in various scenes at a laboratory. The experimental results show that the navigation system can accurately identify the obstacle with an area more than 10-3 m2 and the holes with a height difference more than 0.02 m and can determine the best way to avoid the obstacle in an effective range by voice prompt to the user, so as to meet the daily demand to blind navigation. The development of this system and the applications of various algorithms not only have important effect on the blind navigation, but also have reference significance for the robot walking dependently.