Design of Monocular Vision-based Indoor Positioning System Based on Multi-markers
Wu, Jinping (2017-08-15)
Design of Monocular Vision-based Indoor Positioning System Based on Multi-markers
Wu, Jinping
(15.08.2017)
Tätä artikkelia/julkaisua ei ole tallennettu UTUPubiin. Julkaisun tiedoissa voi kuitenkin olla linkki toisaalle tallennettuun artikkeliin / julkaisuun.
Turun yliopisto
Kuvaus
Siirretty Doriasta
Tiivistelmä
Computer vision has attracted increasing attention on the strength of its sufficient environmental information. In recent years, vision-based indoor positioning has been actively studied by various researches. Among existing vision-based positioning methods, monocular vision-based positioning is the core method for autonomous positioning. However, this approach suffers from three main technology difficulties for its high requirements of indoor positioning: estimating the position and orientation of moving targets accurately, improving the robustness of this positioning system,simplifying the positioning algorithm.
This thesis proposes a low cost vision-based indoor positioning method for the unknown environment with low computation complexity. The framework of the proposed monocular vision-based positioning system starts from marker detection and the information extraction of marker corner. The system detects unique ID of each marker in real time with excellent robustness. The relation between the camera and targets are described by Affine Theorem. The holography matrix and the coordinates of the markers related to the camera can be calculated by establishing the constraint equation. The global coordinate of the camera under the environment coordinate system will be obtained by the change-of-coordinate and change-of-coordinate system recursively.
To measure the accuracy of this positioning algorithm, benchmark tests were carried out on the relative coordinate of markers and the global coordinate of the camera, respectively. The result indicates the proposed system performs well in real time and accuracy but with low cost.
This thesis proposes a low cost vision-based indoor positioning method for the unknown environment with low computation complexity. The framework of the proposed monocular vision-based positioning system starts from marker detection and the information extraction of marker corner. The system detects unique ID of each marker in real time with excellent robustness. The relation between the camera and targets are described by Affine Theorem. The holography matrix and the coordinates of the markers related to the camera can be calculated by establishing the constraint equation. The global coordinate of the camera under the environment coordinate system will be obtained by the change-of-coordinate and change-of-coordinate system recursively.
To measure the accuracy of this positioning algorithm, benchmark tests were carried out on the relative coordinate of markers and the global coordinate of the camera, respectively. The result indicates the proposed system performs well in real time and accuracy but with low cost.