Indoor Scene Positioning Method Based On WiFi Location Fingerprint

The vigorous development of information technology,artificial intelligence and the Internet of Things technology has opened up barriers to the real world and information space.Among them,location services based on indoor scenes have become indispensable Io T perception information in daily life.In this thesis,an indoor scene positioning system based on WiFi is taken as a research object,the WiFi location fingerprint positioning technology has been researched.The key problems in the positioning process are analyzed and the corresponding solutions are then put forward.The main work of this thesis are as follows:1)For the problem of the missing and difference of RSS(Received Signal Strength)signals measured by heterogeneous devices due to different hardware configurations,a gradient normalization algorithm based on missing value compensation is proposed.The WiFi signal characteristics and attenuation law in indoor scenes are analyzed.For the lack of RSS measurement caused by device differences,offline calibration is performed.Difference compensation is carried out for terminal equipment based on gradient invariance,and the effectiveness of the proposed algorithm is verified by experiments.2)For the problem of redundant AP selection with weak representation ability,an AP selection algorithm based on stable information gain is proposed.The characterization capabilities of different APs are analyzed,redundant APs with weak characterization capabilities are filtered,and a subset of APs that are most valuable for positioning are extracted.The effect of the AP selection algorithm on the positioning accuracy under different numbers of APs is verified by experimental tests,and the appropriate number of parameters are selected.3)For the problem of the time-varying and random characteristics of RSS sequences,a hybrid filtering algorithm based on isolation forest and weighted mean is proposed.To solve the problem of delay dislocation caused by the long processing time of RSS in static filtering method,firstly the outliers are eliminated by using the detection algorithm based on isolation forest,and then the RSS after singular value correction is processed by using the weighted average method,thus realizing real-time filtering correction of RSS,ensuring the stability and reliability of RSS,and providing high-quality fingerprint signals for clustering analysis.4)For the problem of the high computational complexity of online location,a Mixed Distance-based Affinity Propagation(MDAP)clustering algorithm is proposed.The hybrid distance which can better reflect the similarity of RSS signals is combined with the affinity propagation clustering algorithm,and damping factor λ is added to accelerate the algorithm convergence,thus reducing the calculation cost of online positioning and improving the real-time and accuracy of positioning.5)The design and implementation of the WiFi indoor scene positioning system based on the Android platform is completed,with an average positioning error of 2.4m in daily office scenes,and the feasibility and effectiveness of the system are verified by experimental environment tests.