This paper designs and implements a visual Simultaneous Localization And Mapping (SLAM) algorithm based on embedded platform and Kinect sensors. Kinect sensors consist of a RGB camera and an infrared CMOS camera that uses structured light to measure depth. The algorithm uses the Oriented FAST and Rotated BRIEF (ORB) operator as the description information of the environmental feature points and uses the nearest neighbor repairing method based on edge detection to correct the depth image so as to obtain a complete depth map. Based on this, Locality-Sensitive Hashing (LSH) algorithm is used to match feature points. Experimental results show that the visual SLAM algorithm based on ORB features is feasible, has good positioning accuracy, and can be widely used in autonomous navigation tasks of indoor robots.