Research and implementation of 3D SLAM algorithm based on Kinect depth sensor

For the 3D perception problem of mobile robots in unknown indoor environments, a practical approach to building 3D maps using a low-cost Kinect sensor is proposed. Successive frames of RGB-D information are captured during the robots movements. First, SIFT(Scale-invariant Feature Transform)detector is applied to color images for extracting and matching stable feature points. Then, combined with the GTM(Graph Transformation Matching) algorithm to eliminate the possible existence of false matching points and complete the initial registration, so as to estimate the rough relative transfer between the image frames. Finally, using the ICP(Iterative Closest Point) algorithm, the robot's motion parameters are continuously updated to update the robot posture, and based on this, the 3D map of the indoor environment is created based on the combination of depth information. Experimental results validate the feasibility and effectiveness of the approach