Study On Indoor Positioning Method Based On Multi-sensors Integration

With the continuous innovation of science and technology,the economic and social has developed vigorously,and the concepts of "digital earth" and "smart city"have been widely promoted and become common knowledge,it makes higher requirements for the speed and quality of indoor high-precision positioning,navigation,and time service based on LBS(Location Based Services).Traditional navigation and positioning techniques has been unable to meet the diverse demands of LBS.The integrated navigation and positioning system based on multi-sensor technology is used to obtain real-time spatial information and realize high-precision,efficient and stable positioning of indoor complex environment,it has become a new research trend.The research purpose of this dissertation is to achieve indoor high-precision navigation and positioning.Based on the existing technology of wireless sensor network,INS,LiDAR and visual navigation and positioning technology,efforts are focus on the theory,method and technology of ZigBee indoor positioning system,high-precision visual positioning method,INS/LiDAR/Vision technology integrated positioning and fusion mechanism.Some key technologies like wireless signal path loss model,high-precision spatio-temporal information registration,robust and efficient fusion algorithm and accuracy evaluation will be studied in this paper.The main contributions obtained in this paper are as follows:1.This paper introduces the protocol of ZigBee wireless sensor network technology,studies the positioning method of path loss model based on RSSI(Received Signal Strength Indication).Aiming at the problem that the parameters in this model are difficult to determine,a method for determining model parameters based on PSO(Particle Swarm Optimization)algorithm is proposed,and a two-dimensional positioning method based on PSO algorithm and kNN(k-Nearest Neighbor)algorithm is designed.A ZigBee wireless sensor network is builded in the indoor environment,and positioning experiments are performed to verify the effectiveness of the proposed algorithm and positioning system.2.Aiming at the requirement for stable and high-precision positioning of robots in indoor environments,an integrated navigation and positioning algorithm based on vision and inertial sensors is proposed.For the positioning algorithm based on Kinect vision,SIFT(Scale-invariant feature transform,SIFT)algorithm is used for feature point extraction and matching,and RANSAC(RANdom SAmple Consensus)algorithm and bubble method are used to obtain more robust feature points.Finally,an absolute orientation algorithm is used to calculate the translation.This method has better positioning accuracy than the traditional ICP algorithm.For the inertial navigation system,we analyze the equations of its position,velocity and attitude.This paper designs and implements a Kinect vision/inertial integrated navigation and positioning system based on Kalman filtering.Through indoor experimental tests,it is verified that the integrated navigation and positioning system can effectively achieve indoor positioning.Compared with the pure vision positioning method,the designed integrated navigation and positioning system improves the positioning accuracy and reliability.3.In order to further improve the indoor positioning accuracy and stability of the robot,an INS/LiDAR/vision integrated indoor positioning method based on federated filtering is proposed.For the processing of 3D point cloud data of LiDAR,ISS feature extraction algorithm and a feature extraction algorithm based on the voxel-scale invariant feature transform(Voxel-SIFT feature)proposed by the research group are studied and compared.Voxel-SIFT feature extraction algorithm with a large number of extracted feature points and a short processing time is selected as the feature extraction algorithm in the subsequent registration stage.Based on this,the Kd-Tree algorithm and ICP registration algorithm are used to complete the accurate registration of point cloud data.This registration algorithm takes significantly less time than the traditional ICP registration algorithm.For the selection of the visual odometer front end,the feature-based method and the direct method of visual odometer algorithm were studied,through the experiment,the DSO(Direct Sparse Odometer)algorithm is selected as the visual front-end.Finally,on the basis of the traditional federated filtering,a least squares iterative weighting algorithm is designed to fuse the optimal estimates of the sub-filters.The INS,LiDAR and vision integrated navigation and positioning system based on the federated filtering algorithm is implemented.Through the indoor scene experiment,it proves that the system improve the stability of the positioning system and can achieve high-precision indoor positioning.