Research On Two Dimensional Target Registration And Localization Algorithm And System Based On The Mobile Terminal

Augmented Reality(AR) technology is one of the hot topics in the area of computer applications. AR technology can merge virtual objects with the real scenes to enhance the user's understanding and experience of the real world. Mobile Augmented Reality(MAR) refers to the augmented reality systems running on mobile devices such as smart phones, portable computers, etc. More conventional augmented reality systems use desktops and workstations as the operating platform, and such systems limit the scope of the user and cannot be applied to the outdoor environment. The rapid development of mobile devices and network technology releases augmented reality technology from the restrictions of PC, workstation and other heavy equipment, and promotes the emergence and development of MAR.An augmented reality system is composed of modules such as virtual graphics generation?registration and localization?virtual-real and display. Currently there still exists a gap between the performance of the mobile terminal and that of the general personal computer, that the registration and localization algorithms designed for the personal computer platform are not applicable. This thesis focuses on the research of registration and localization technology of two dimensional target based on mobile terminal. The main research contents are listed as the followings:1 Research on the augmented reality technology and mobile augmented reality background,application field, domestic and foreign development situation of registration and localization technology of two-dimension objects present as well as the realization method. The key techniques of the registration are briefly analyzed, including image feature detection, feature matching, feature tracking and pose estimation. And pointed out that the focus on the research is registration and localization technology of two dimensional target based on mobile terminal.2. In the third chapter, the design and construction of a 2D target registration and localization system framework is designed and built, for the characteristics of the weak and the hardware resource of the mobile terminal, including image collection, image tracking, image matching, fusion module, pose estimation, virtual reality fusion and enhanced display modules. Then a novel and fast matching method which combines the tracking and matching is proposed for the performance and functional requirements of the system matching module and tracking module. The main idea is to let binary descriptor matching and optical flow tracking run in parallel and real-time interaction according to the area of overlap ratio and using two-way FLANN and RANSAC algorithm to preliminary refinement for matching and tracking, and put forward the four strategies to further refinement. Experiments show that compared with using features matching alone or tracking, the method effectively improves the registration and the tracking the real-time and accuracy.3.On the basis of the matching tracking, the camera pose estimation and virtual-real technology are studied in the paper, and we introduced the principle and design of the pose estimation module and virtual reality fusion and enhanced display modules. And using geometric constraint to improve RANSAC algorithm, and to improve the speed of homography operations. By adopting the improved sovle Pn P algorithm can get the camera pose without camera intrinsic parameters,and the application scenario of the proposed algorithm is extended. The thesis proposes a fast loading method of 3d model files to accelerate the speed of loading model files.4.The thesis has completed a 2d target registration and localization prototype system. The system is developed by using jni(Java Native Interface) in android platform.Models built by 3ds Max software are displayed with real-time on the phone screen by using Open GL ES graphics rendering tools according to the camera pose. Don't need to be set in advance, the system can select the interested target as the target template. At the same time, the system can be realized to robust registration and tracking in the current mainstream mobile platforms, at a speed of above 25 FPS. Experiments show that the prototype system has good robustness and real-time performance in the normal, scale, scaling, rotation and partial occlusion.