Research On Indoor Visible Light Localization Based On Fingerprint Algorithm

Indoor visible light positioning as a new indoor positioning solution,compared with the traditional positioning method,positioning accuracy is high,low cost,no electromagnetic interference,high security and rich spectrum resources,with these advantages of visible light in indoor positioning has great potential and research value.In this paper,for the fingerprint algorithm of indoor visible light positioning,there is a large workload in the process of constructing fingerprint database in the offline stage and it needs to be reconstructed when the scene changes.In addition,the Weighted K Nearest Neighbors(WKNN)algorithm used in the online positioning stage has problems such as low positioning accuracy and long positioning time.To improve the traditional fingerprint algorithm,the main contents of this paper are as follows:In the process of offline fingerprint database construction,a Lambert propagation model of optical signals is proposed to measure the intensity of optical signals in a few locations,and to calculate the unknown values of other locations through the known measured values,so as to construct a virtual fingerprint database.In the online positioning stage,an improved WKNN algorithm is proposed to study the influence of the selection of K value on the positioning accuracy,and the self-adaptive changing K value is used to replace the fixed K value.Comparing the positioning accuracy under different distance measurements,the square chordal distance measurement is used to replace the Euclidean distance measurement.The locating area of the target is divided according to the light power.The method of locating step by step is adopted to determine the approximate area of the target first and then match it with the fingerprint database in the region,which reduces the workload of matching with the whole fingerprint database and reduces the locating time.Considering the influence of the stability of Light Emitting Diode(LED)on the positioning system,the variance of LED signal intensity is introduced into the calculation of K value weight to improve the positioning accuracy.Finally,a real visible light positioning system is built,and the intensity of optical signals is collected to verify the feasibility of the algorithm proposed in this paper.After experimental simulation and building a real positioning system for verification,the results show that the virtual fingerprint database constructed by optical signal propagation model is more accurate than the virtual fingerprint database constructed by Kriging interpolation and matrix filling,and it also effectively simplifies the construction process of the fingerprint database.Compared with the compressed sensing and extreme learning machine positioning algorithm,the improved WKNN positioning algorithm can achieve higher positioning accuracy,solve the problem of large positioning error in the edge region,and greatly reduce the positioning time,thus achieving the research purpose of this paper.Figure[52]Table[5]Reference[84]...