基於單張影像後方交會的手機即時室內定位系統

An astonishing growth of smartphone usage and real time navigation popularity have been witnessed in recent years. Hence, Location Based Service (LBS) has become increasingly important for a rising number of applications in different field. In outdoor environments, the Global Navigation Satellite System (GNSS) can be exploited in conjunction with Inertial Navigation Systems (INS) and some algorithms of map constraint to achieve the goal for obtaining the precise position information. However, in indoor environments, satellite signal cannot be received due to the shadowing problem, so that indoor positioning technique should seek out other technologies to get location information. Consequently, various indoor positioning technologies began to develop and lots of service are provided nowadays. Particularly, vision-based technology has main advantages including the infrastructure is not necessary and the accuracy has reached appreciable levels. If the time burden of image processing can be improved, vision-based has a good prospect since the growth of computational capabilities, image processing, and the popularity of smartphone built-in camera. Therefore, the research designs an application on mobile phone, not only attaches markers on object to decrease processing time, but also does camera calibration to improve the accuracy in advance. In addition, the research develops an algorithm assisted by Canny and contour detection algorithm of OpenCV library to find markers effectively. In offline stage, the application has constructed a database of control points in object coordinate system and the coefficient of Interior Orientation Parameters (IOPs). Under these circumstances, user just takes one photo then the application can detect those markers and doing ORB matching with RANSAC to retrieve those attributes according objects from database. The application can successfully match objects from database and through space resection to achieve indoor positioning in centimeter-level. Furthermore, vision-based positioning can calibrate the error which increases over time and produced by PDR and take their respective advantages to achieve rapid and accurate indoor positioning technology. Overall, the indoor positioning technology which is proposed by the research is a low cost, high processing speed and high accuracy real time method.