Abstarct: In this thesis, we have developed a 3D reconstruction system baxsed on RGB-D data. The proposed approach is suitable for both 3D object modeling and indoor reconstruction. First, we record the RGB-D data by a Kinect sensor that is held by hand. The depth of the sensor is refined in the pre-processing stage and its noise is partially eliminated. The external parameters of the camera are obtained in motion between two frxames (including the rotational matrix and displacement vector), and then track the motion of the camera using a using a robust 3D alignment algorithm baxsed on RANSAC. Then, we check the camera movement rate to increase the accuracy of mapping. The basic map of the environment, which includes camera poses between the two frxames, features and observations of the scene, is made using frxame tracking to the frxame and added to the overall map. In order to reduce memory consumption and improve processing time, instead of storing information from all 3D cloud pixels, we use voxels and reduce the cloud-point volume. Finally, the kinect sensor for each pixel gives the exact depth, the indoor 3D map is an in-depth metric scale, which can be useful for in-depth visual inspections, reverse engineering and measurement processes.