基于Kinect的障碍物探测及三维场景重建研究与实现-信号与信息处理专业论文.docxVIP

Subject : Research and implementation of obstacle detection and 3D scene reconstruction based on Kinect Specialty : Signal and information processing Name : Zhao Liang (Signature) Instructor : Sun Yi (Signature) ABSTRACT Reliable depth data is a cornerstone of many autonomous robotic control systems. Microsoft Kinect Has a wide range of applications in recent years in the rise of 3D printing, AR and VR as a Low cost, small size, high precision depth data of the sensor. Because of the cheap equipment and the output depth data is stable and reliable while the application which use depth data based on Kinect became popular in the field of 3D modeling and robot navigation. According to the different needs a number of targeted solution had been putted forward by scholars at home and abroad, also some fruits are achieved. The depth data processing of Kinect is aims mainly in this article, the method of obstacle detection and 3D scene reconstruction which based on depth data of Kinect is also proposed. Equipped with Kinect depth sensor by quad rotor, the sensor can collect the depth data in real time that the scene have been tagged, then upload to PC for obstacle detection and 3D scene reconstruction. This depth sensor use transmitting laser speckle as input quantity and depth data which calculating by computer chip as output. The sensor upload depth data to the host computer in the process of scanning and then host computer change the depth data to 3D point cloud. Through the camera tracking, point cloud fusion, scene rendering process which change the depth data to the 3D scene of .obj format. Because the 3D scene reconstruction requires high hardware conditions which based on Kinect, only compatible with the latest DirectX 11 GPU to support real-time, interactive reconstruction. So in the construction of 3D scene, using high performance GPU to process data to insure the accuracy of 3D scene. Proved through the experiment Kinect can detect obstacles and display 3D