Research On Vision-based Localization Using Pixel Threshold Based Eight Points Algorithm

With the rapid development of science and technology,localization technology has become a hot issue among both industry and academia.GPS based localization technology has been mature in the outdoor environment.However,GPS signal is not suitable for indoor positioning because of the interference from walls and pedestrians.In recent years,the indoor vision-based localization system has arisen due to the popularity of smart phones,which has low cost and good localization performance and is easy to promote.By summarizing the characteristics of several indoor localization technologies,this paper analyzes the purpose and significance of indoor vision-based localization and studies the research status of indoor localization technology and indoor vision-based localization technology.Besides,some theoretical knowledge related to the indoor localization system is introduced.The work has been finished is shown as follow:In view of the fact that the existing indoor vision-based localization system has a large amount of information in the database and the image retrieval algorithm is time consuming,a fast image retrieval algorithm based on image classification is proposed in this paper.A visual histogram construction method based on image segmentation is adopted to improve the accuracy of image representation.And support vector machine algorithm is used for supervised training to get the classifier.In the online phase,the user image is classified and performs image retrieval in a smaller range according to the label.The simulation results show that the proposed algorithm effectively improves the accuracy of the classification and reduces the image retrieval algorithm complexity in order to improve the time performance of the positioning system.Aiming at the poor robustness of the eight points algorithm during calculating fundamental matrix,an improved eight points algorithm based on pixel threshold is proposed in this paper to calculate fundamental matrix.It introduces pixel distance as a new criterion for selecting eight pairs of matching points and eliminate the distorted matching points caused by environmental changes.Simulation results show that the improved eight points algorithm based on pixel threshold has better performance during calculating the fundamental matrix.It can obviously reduce the fundamental matrix error and improve the accuracy of the localization system.