Research On Indoor Location Method Of Multi-source Data Fusion Based On Behavior Perception Assistance

With the development of Intelligent Transportation applications,the pedestrian positioning navigation system came into being,and the positioning technology is the core and foundation of the positioning navigation system.At present,the outdoor positioning system is well developed and the GPS satellite positioning is accurate.However,due to wall occlusion in the indoor environment,the GPS signal is weak and cannot be used for indoor positioning.Therefore,this paper is devoted to the study of low-cost,high-precision indoor positioning technology.Nowadays,in the mainstream indoor positioning technology,most positioning technologies require additional positioning of base stations or tags,and the positioning cost is high.Smartphone-based Wi-Fi positioning and Pedestrian Dead Reckoning(PDR)are easy to promote and popular without additional equipment.However,regardless of Wi-Fi or PDR,the positioning accuracy of individual positioning sources is limited.Among them,the uncertainty of Wi-Fi wireless signal strength and the multipath effect of signal propagation result in poor accuracy and stability of Wi-Fi positioning.However,PDR has accumulated error problems and cannot be located for a long time.As smartphone sensors become more powerful,smartphones have strong behavioral awareness.Based on this,this paper proposes a multi-source data fusion indoor positioning method based on behavior-aware assistance.The specific research content is as follows.Firstly,a behavior-aware model based on deep learning is studied.It mainly uses mobile phone sensors to collect various behavior data in the room,including static,straight,turn,stairs,elevator and escalator.Through the deep learning TensorFlow framework and convolutional neural network,the behavioral awareness model is trained and transplanted to the Android platform to realize online behavior recognition.Then the main functions of behavior-aware assisted positioning are proposed,including positional reasoning based on behavior recognition and adaptive learning of positioning parameters.Then,some improvements are proposed for the positioning methods of Wi-Fi and PDR,including Wi-Fi positioning adaptive KNN method and PDR parameter adaptive learning process.Finally,a fusion positioning model is proposed,which combines the above three positioning technologies to achieve multi-source data fusion positioning.In addition,the fusion positioning model proposed in this paper is more scalable,and other positioning sources can be directly added in subsequent research.Based on the above research content,this paper develops an indoor positioning APP based on the fusion positioning algorithm with the smartphone as the terminal and the Android system as the platform.The statistical analysis of the experimental data was successfully completed with the aid of the positioning system.The experimental results show that the fusion positioning algorithm has an average positioning accuracy of 1.3 meters and a positioning standard deviation of about 0.75 meters,which effectively improves the accuracy and stability of positioning.In addition,in the National Indoor Positioning and Navigation Competition organized by the Ministry of Science and Technology of China in November 2018,the method won the champion of “No Smartphone Supported by Additional Equipment” and verified the effectiveness of the algorithm.