Fiducial marker indoor localization with Artificial Neural Network

A vision based positioning system could be categorized into two groups. One analyzes an environment's scenery by matching the inputs with imaginary database to find the optimum result. The other uses fiduciary markers. In proposed method, the system uses fiduciary markers with a capital alphabet in it. When the known size fiduciary marker is captured by a camera, by using homography transformation, the 6-DOF camera pose with respect to the marker's local coordinate can be calculated. To recognize the character in the marker, Artificial Neural Network (ANN) with back-propagation training method is used. 12 unique features of a character are defined and used as inputs of ANN. Since more than 95% recognition rate is achieved in testing phase, the Optical Character Recognition (OCR) with ANN could be used as a marker detection method. The localization experimental result with the fiduciary marker shows that the proposed method could be a solution for indoor localization.