| The "smart world" envisioned by the Internet of Things is realized by infiltrating intelligence into ubiquitous things in life,including physical objects,network entities,and social elements.With the advancement of science and technology,wireless positioning technology has been widely used in society.Among them,the device-free localization(DFL)technology of the target based on wireless signals,because it does not need to carry any electronic equipment for positioning on the located target,it is used in anti-intrusion detection and smart home construction,etc.It has advantages in application and has become a current research hotspot.Aiming at the problems of low indoor positioning accuracy and unstable positioning performance caused by radio signals susceptible to environmental interference,as well as positioning calibration and positioning scene mobility in the DFL positioning process,the following research is carried out in this paper.1.This paper builds a new sensor network configuration mode,which can use a small number of sensors to obtain high positioning accuracy.In this mode,by moving the transmit antenna,only a few receiving sensors are required to achieve sufficient wireless links in the sensor network,resulting in excellent localization performance.This paper uses Universal Software Radio Peripheral(USRP)to build a wireless sensor positioning system in the laboratory environment,and uses the Received Signal Strength(RSS)of the wireless signal as the positioning feature to locate the target.2.In this paper,the DFL problem is transformed into a sparse representation problem,and the positioning accuracy is improved by improving the performance of the algorithm.When constructing the objective equation,the GMC(Generalized Minimax-concave,GMC)regularization function is introduced as the relaxation of the L0 norm.The Forward-backward Splitting algorithm(FBS)is used to optimize the objective function to obtain the global optimal solution,thereby improving the accuracy of target positioning.3.In this paper,an Adaptive Relaxation Localization(ARL)algorithm for target position prediction is designed,which treats the localization problem as a regression problem and significantly improves the localization performance.The experimental results show that 100% localization accuracy can be achieved at a position resolution of 0.5 m ×0.5 m,and it has good anti-noise performance.The average positioning error is only 0.053 meters at a position resolution of 0.25 m × 0.25 m,which is the best positioning performance in the DFL field.4.In order to further improve the positioning accuracy and solve the migration problem of the positioning algorithm,this paper adopts a label-consistent supervised dictionary learning method to solve the DFL problem.The method was first validated on the public localization dataset of the University of Utah,and achieved excellent performance.Then,the positioning data was collected in the laboratory positioning system for positioning experiments,and the positioning performance was compared with various algorithms to verify the effectiveness and robustness of the supervised dictionary learning method. |
