Research On Indoor Vision Based Localization Using Improved Ransac Algorithm

With the rapid development of science and technology,users' demand for their location information in indoor environment is growing.The traditional outdoor localization signal is usually difficult to be obtained in indoor environment.Thus,the researchers have proposed a vision-based indoor localization method.This paper studies the research status about indoor localization and vision based indoor localization.Besides,some related theoretical knowledge are introduced.The work we finished is as follows:Aiming at the problem that the exiting vision-based is time-consuming in the stage of image retrieval,a supervised classification-based fast image retrieval algorithm has been proposed.In order to increase the accuracy of indoor localization,the scale of image database should be as large as possible.However,this could course huge time consuming.Thus,this paper utilizes SVM supervised classification algorithm to classify the query image.The simulation results show that the supervised classification-based fast image retrieval algorithm can decrease the time consuming of image retrieval process.Traditional RANSAC algorithm is lack of stability,whose iterations are not fixed and time consuming is often huge.This paper proposes the improved RANSAC algorithm.Because RANSAC algorithm chooses samples randomly to estimate the model parameter,which may not be inners.In this paper,we define a measure factor ? to represent the probability whether a pair of matching points is inners.In the sampling stage,the improved algorithm choose the matching pairs with higher measure factor to reducing the number of iterations and time consuming.When this algorithm is applied to the indoor environment,the result shows that our proposed method achieved approximately time consuming reduction by 67%.Though the traditional method has a little better localization performance than our proposed method,but it will pay much more time.The proposed fast localization method could make balance between accuracy and efficiency.It could be well implemented in the indoor environment for vision-based indoor localization.