|Intelligent shadow matching based on improved multi-classifier for urban positioning
|Hong Kong Polytechnic University -- Dissertations
Global Positioning System
Geographic information systems
|Faculty of Engineering
|viii, 92 pages : color illustrations
|It is well known that the Global Navigation Satellite System (GNSS) performance in urban area is unreliable because the surrounding construction can block and reflect satellite signals. The shadow matching is a positioning method by using GNSS information and three-dimension (3D) building model to solve the problem in urban canyon. For the shadow matching algorithm, the position solution is determined by comparing the received signal visibility with predicted signal visibility over a grid map. In this way, the GNSS line-of-sight (LOS) signal and non-line-of-sight (NLOS) signal classification is very important for shadow matching and other conventional positioning approach. This paper describes an improved classifier algorithm based on multi-classifier for shadow matching in dense urban area. The multi-classifier contains a signal-to-noise ratio (SNR) classifier, a machine learning classifier, and two pseudorange rate consistency classifiers. The machine learning approach is selected after comparing different machine learning methods, including k-nearest neighbors (KNN), neural network (NN), decision tree (DT), and support vector machine (SVM). The features for machine learning approach are 1) Carrier to noise ratio; 2) Satellite elevation; 3) Normalized pseudorange residual; 4) Pseudorange rate consistency. The improved multi-classifier also provides the confidence coefficient of classification for shadow matching scoring. Using Ublox and smart phone data recorded in Hong Kong, it is shown that the best classification accuracy comes up to 96% at Tsim Sha Tsui, 91.8% at Tsuen Wan and 99.7% at Mong Kok by considering the confidence coefficient. The shadow matching with an improved classifier algorithm achieved the best mean position error of 6.77m among the static tests and 7.05m among the pedestrian tests at Tsim Sha Tsui. Compared with the SVM classification, this represents 5~6m improvement in dynamic tests and approximately 10m improvement in static tests. For the Tsuen Wan experiments, there are 2~5m improvement in dynamic tests and around 5~12m improvement in static tests. In addition, the relationship between classification accuracy and shadow matching performance are simulated and discussed, and the challenges of urban classification also are discussed.
|All rights reserved
Files in This Item:
|For All Users (off-campus access for PolyU Staff & Students only)
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item: