Research On Indoor Visible Light Positioning Fusion Algorithm Based On BP Neural Network

With the rapid development of mobile communication networks and electronic information technology,people are paying more and more attention to location based services(LBS).Global positioning system(GPS)is the most famous global navigation satellite system,which can provide users with meter-level location services and is now used in various outdoor location services.However,in the indoor environment,the signal from the satellite is affected by the blockage of buildings and multipath effect,so the indoor positioning accuracy by GPS will be very low,which cannot meet the demand of indoor location services.At present,indoor positioning mostly uses technologies such as Bluetooth,infrared ray,wireless local area network,radio frequency and ultra-wideband,but these positioning technologies generally require the support of complex hardware facilities and are susceptible to electromagnetic interference.Compared with the abovementioned positioning technologies,visible light positioning technology has the advantages of low positioning cost,no electromagnetic interference,high security and high positioning accuracy,and has become one of the hotspots for research scholars at home and abroad.This paper will also focus on indoor visible light positioning technology and achieve high-precision indoor positioning.The main work is as follows:(1)The basic characteristics of the LED light source and the indoor visible light channel model are introduced,and the indoor illuminance distribution and indoor received optical power distribution under line-of-sight link are simulated.The common indoor visible light positioning algorithms are introduced in detail,and the advantages and disadvantages of several common indoor visible light positioning algorithms are compared,and the location fingerprint algorithm is finally selected as the research focus.(2)In order to avoid the inter-cell interference(ICI)caused by multiple LED lamps indoor visible light positioning system during positioning,a novel single LED lamp indoor visible light positioning system is proposed,which uses a single LED lamp as the transmitter,three horizontal photodetectors(PDs)as the receiver,and the point to be measured is located in the center of the receiver.The common location fingerprint algorithms(NN,KNN,WKNN and Bayes)are introduced in detail,and the positioning accuracy of the four positioning algorithms is simulated,among which the WKNN algorithm has the highest positioning accuracy.(3)The location fingerprint algorithm is to match the actual fingerprint of the point to be measured with the fingerprint data of the whole location fingerprint database,which is not only computationally intensive,but also some unnecessary fingerprint data will also affect the positioning results.In order to achieve high-precision indoor positioning,an indoor visible light positioning algorithm based on double BP neural network is proposed.The algorithm first uses the BP neural network to determine the rough location range of the point to be measured,and then uses the location range as a limiting condition to use the BP neural network again to achieve more accurate positioning.The simulation results show that in the indoor space of 2m×2m×2.5m,the average positioning error of the algorithm is 0.16 cm and the average positioning time is 0.48886 s.Compared with the WKNN algorithm,the positioning accuracy of the algorithm is improved by 97.61%,and the positioning time is increased by 98.68%.Therefore,although the double BP neural network algorithm can effectively improve the positioning accuracy,its positioning realtime performance is poor.Then an indoor visible light positioning algorithm based BP neural network and multiple linear regression is proposed,which only uses the multiple linear regression algorithm to replace the BP neural network algorithm in accurate positioning.Under the same simulation environment,the average positioning error of the algorithm is 0.38 cm and the average positioning time is 0.00292 s.Compared with WKNN algorithm,the positioning accuracy of the algorithm is improved by 94.33%,and the positioning time is reduced by 54.73%.Therefore,the BP neural network and multiple linear regression fusion algorithm can not only effectively improve the positioning accuracy,but also its positioning real-time performance is better.(4)The experimental scene of a single LED lamp indoor visible light positioning system is built to compare and analyze the positioning performance of the double BP neural network algorithm and the BP neural network and multiple linear regression fusion algorithm in the actual indoor environment.The experimental results show that in the indoor space of 2m×2m×2.5m,the average positioning errors of the two algorithms are5.99 cm and 5.04 cm,and the average positioning time is 0.47999 s and 0.00283 s,respectively.Therefore,the BP neural network and multiple linear regression fusion algorithm has better positioning performance in the actual indoor environment.