A Weighted Fingerprint Location Algorithm Based On Propagation Model And Clustering

Positioning technology has experienced rapid development in recent years,and location-based services and applications are increasingly being used in daily life.The positioning method is roughly divided into three types: approximation method,geometric method and scene analysis method.The scene analysis method is commonly used.The scene analysis method locates the object according to the specific scene information collected by the object to be measured.This paper uses the location fingerprint based method for positioning.During the offline fingerprint database establishment phase,the features in the scene are detected.The wireless signal is detected in this paper.In the selection of wireless signals.Although Wi-Fi signals cover a wide range,they need to continuously scan the device when positioning,and the power consumption is relatively large.Therefore,in this paper,we use ZigBee devices with relatively low power consumption as devices for transmitting and receiving wireless signals.After the comparison,the CC2530 chip introduced by TI was selected,and the ZigBee wireless sensor network was built based on the Z-stack protocol stack to collect the wireless signals.In the online positioning stage,the RSS model strength information collected according to the object to be located is matched with the offline fingerprint database constructed in the offline phase.The commonly used method is the weighted k-nearest neighbor(WKNN)algorithm,which firstly needs to select k points in the offline fingerprint database.In this paper,the K-means clustering algorithm is chosen,but the Kmeans clustering algorithm randomly selects the initial center,which causes different choices to affect the positioning results.The K-means clustering algorithm has been improved to select nodes that are relatively far apart from each other as the initial center.Then,in the choice of weight value,the traditional method is to use the reciprocal of Euclidean distance as the weight value.However,the Euclidean distance of the RSS signal and the actual physical distance are not a simple linear relationship,which can be expressed by a physical propagation model.Therefore,through the actual environment test and experimental simulation,this paper uses the combination of physical propagation model and traditional Manhattan distance to select the weight value,uses fingerprint location method to complete the positioning,and analyzes the positioning accuracy and comparison.The effect of different parameters on the results of positioning accuracy.