In recent years, the number of mobile robot has increased rapidly. They can not only be used for carrying out tasks in hostile work environment, but also make people's daily life convenient. In the meantime, the rise of 3D technology has changed people's life with its applications in the fields of industrial life and daily life. Therefore, the combination of 3D technology and mobile robot has enormous potentials.

The objective of this project is to create a model of an object or indoor environment with the help of a mobile robot. A 3D scanner which can be attached on a mobile robot will be used for scanning. With the depth information collected, the point clouds can be gotten. The basic way of reconstructing the model is called Kinect Fusion, it uses Kinect camera to scan the area and creates a model with help of Iterative Closest Point algorithm and Truncated signed distance function which will be introduced in section 2.5. The registration of the point clouds and the localization of the camera can be realized by using ICP algorithm. The software which is used to implement these methods is Visual Studio 2012 with OpenCV and KinectFusion library.

As the camera should be able to be used on any kind of robot, the aim of these tests is to find a better way to get a high quality model of the area. In this paper, two main experiments have been done. One of them is about scanning a target object from different distances. There will be errors between object actual size and its model size. And the relationship between the errors and the distances will be found out. The other one is focused on the movement of the camera. There are two scenarios in this experiment. The first one is to move the camera with different speed and the other scenario is to scan an object by moving the camera around the object. The result shows that the robot attached with the camera should move slowly so that the camera can capture enough frame images for the model. And keep scanning one object by moving the camera around it can't make the model looks better but worse.