Research On Indoor Location Method Based On Clustering And Neural Network

With the rapid development of indoor positioning technology,people's demand for indoor positioning in many scenarios is also increasing day by day.In the actual positioning situation,due to the unstable environment,the indoor positioning results will be affected.Therefore,improving positioning efficiency and accuracy has become the direction of indoor positioning research.Based on the existing indoor positioning principles and methods,this paper has carried out relevant research on indoor positioning methods based on signal strength.The research contents include the following three aspects:(1)An improved indoor location algorithm based on clustering is proposed.The fingerprint database is built by the improved clustering algorithm in the room.An experimental platform is set up,and fingerprint library is built.The initial value is calculated by k-means algorithm,and then the clustering fingerprint library is obtained by algorithm.(2)A combination of Kalman filtering and improved clustering method is used in the room.Calculate the Euclidean distance between each cluster center and the test signal,and divide the test points into the classes with the minimum Euclidean distance value until all test points are divided.The coordinates of corresponding points are estimated by fingerprint location.The physical position of the measured points is obtained by Kalman filtering,which improves the final positioning accuracy.(3)Based on the research of RBF neural network and LANDMARC localization method,an improved LANDMARC localization method based on RBF neural network was proposed.Study first to filter the signal,the signal filtering of the singular point,and then by using density clustering algorithm for noise points and boundary points related to processing,finally through the application of neural network,establish the input and output of the nonlinear mapping relation between positioning error,and will stay input samples of received signal strength and initial positioning coordinates of measuring points to artificial neural network training model and implement the localization,so as to improve the indoor positioning accuracy.Through the experimental simulation,it can be seen that the average positioning error of the improved clustering algorithm can reach 1.651 m.The improved positioning algorithm based on RBF neural network has an average positioning error of 0.931 m.The simulation results prove the feasibility of the proposed method,verify the excellent performance of the improved indoor joint positioning algorithm,improve the accuracy of indoor positioning algorithm,so as to lay a technical foundation for the study of indoor positioning technology.