A Study Of LBS-oriented Machine Vision Aided Inertial Navigation On Smartphones

In recent years, with the continuous development of smart phones and mobile Internet, LBS(Local Based Services) technology has attracted more and more attention. Inertial navigation technology, as a supplement of localization, is important in LBS. Inertial navigation system uses the data measured by the inertial measurement unit and then calculates the posture and position information of the vehicle by the method of filtering and integral, in the aim of positioning the vehicle. The inertial navigation of the smart mobile devices provides technical support for the intelligent robot, indoor navigation, augmented reality and interactive entertainment and other fields. Inertial components of smart phones, however, due to its own reasons, tend to have some big deviation and uncertainty of the error factors. In this case, the navigation drift phenomenon is serious and the navigation precision is low. For these reasons, inertial navigation can’t be applied on smartphones.This paper proposes a computer vision aided inertial navigation algorithm to solve the problem listed above. In this paper, SFM problem solving algorithm is used for threedimensional reconstruction of feature points and the GPU acceleration computing technology on smartphone is applied to track those points in real time. While feature points matches are used by Pn P problem solve algorithm to compute posture and position changes to correct results of inertial navigation. The experimental results show that the proposed algorithm can reduce the inertial navigation drift and improve the accuracy of inertial navigation. In this paper, the main work includes:1. The inertial components and their working principle are introduced. The traditional inertial navigation system and the corresponding algorithm are also introduced. The inertial components on the smart phone(accelerometer and gyroscope and magnetometers) are introduced, and the inertial navigation error analysis on for smart phones is given.2. SFM problem and Pn P problem in computer vision and their theoretical basis are introduced in detail. The proposed algorithm in this paper firstly extracts the feature points of the two-view images and tracks them by pyramids Lucas-Kanade optical flow algorithm. In the Pn P problem solving algorithm, posture and position changes can becalculated by those feature points matches.3. From the angle of theory, Kalman filter algorithm is introduced in detail. Then Kalman filter model used in this paper is given to fusion inertial navigation data and computer vision matching data.4. The Open GL ES 2.0 library are introduced in detail, and by using the Open GL ES 2.0 GPU- accelerated Pyramid Lucas-Kanade optical flow algorithm on i OS platform is implemented to ensure the real-time performance of the entire computer vision aided inertial navigation algorithm.5. The proposed computer vision aided inertial navigation algorithm is applied on different smartphones for testing. In the end, comparison results and error analysis is listed.