According to the research needs of the co-pilot of the aircraft cockpit, in order to enable the robotic arm to complete the navigation task in the aircraft cockpit, this paper aims to construct a 3D map of the simulated cockpit of the aircraft that can be used for the robotic arm navigation. In view of the influence of the distribution of feature points on the accuracy of SLAM( simultaneous localization and mapping) mapping, through experimental comparison, this paper verifies that the improved ORB( oriented FAST and rotated BRIEF) feature detection algorithm of ORB-SLAM is more uniform than the feature points detected and extracted by the SIFT, SURF and ORB algorithms in the OpenCV library, which is more suitable for SLAM. In this paper, Kinect V2. 0 is used as the data acquisition device for depth information images and color images. In the real environment of the simulated cockpit of the aircraft, combined with the ROS system and the ORBSLAM2 system framework, a 3D dense point cloud map in the aircraft cockpit is constructed. In view of the problems of large data and difficult to achieve navigation in point cloud maps, the OctoMap data model is used, which can compress point clouds, adjust resolution and determine whether the space is occupied, convert the point cloud map into an octree map, and finally obtain a 3D octree map with small data and suitable for navigation.