Research On High Precision Visible Light Positioning Technology Based On Machine Vision

The development of internet of things technology and the popularity of mobile intelligent devices makes people’s demand for wireless location services gradually increase.In outdoor scenes,Global Navigation Satellite System(GNSS)has become the main way for people to obtain location services.In recent years,as more and more production and living activities are completed indoors,the indoor location based service(ILBS)has become an indispensable part of the indoor scene.However,GNSS cannot achieve acceptable results indoors because the satellite signal can be weakened by buildings and walls.In order to meet the requirements of indoor high-precision positioning,a variety of indoor positioning solutions based on wireless communication technology have been proposed,including WiFi,Bluetooth,radio frequency identification(RFID),infrared ray(IR),ultra wide band(UWB),ultrasonic,cellular mobile network,inertial navigation and geomagnetic based positioning methods.Compared with the above indoor positioning technologies,visible light positioning(VLP)technology has attracted extensive attention because of its advantages of high security,good confidentiality,low cost,no electromagnetic interference,good positioning effect and so on.But so far,most studies of VLP technology focused on the active navigation positioning system based on ontology,less attention has been paid to VLP technology compatible with security monitoring systems.Moreover,most VLP systems require multiple LED lights at the transmitter,which leads to poor scalability.In addition,the positioning performance of some VLP systems will be affected when there is a relative tilt between the transmitter and the receiver.Aiming at the above problems,a single LED anti-tilt VLP system compatible with security monitoring system is proposed.The main research contents and innovations of this paper are as follows:(1)In this paper,a single LED VLP system based on machine vision is designed to solve the problems of the existing VLP system,such as incompatibility with security monitoring system,large number of LED at the transmitter end and poor anti-tilt performance.The system can be compatible with security monitoring system,and the positioning algorithm based on deep neural network(DNN)is adopted to achieve high-precision anti-tilt positioning.(2)A single LED VLP experimental platform based on machine vision is designed and tested.The experimental results show that when there is no relative tilt between sender and receiver,the average positioning error and maximum positioning error of the system are 0.585cm and 1.83cm respectively,and the positioning error of about 90%of the test points is less than 1cm.When there is a relative tilt angle of 30° between the transmitter and the receiver,the average positioning error of the system is 1.174cm,that is,the centimeter level indoor accurate positioning can still be realized.The average positioning time of the system is 10.72ms,which can meet the requirements of real-time positioning.(3)A data sparseness strategy was designed to verify the influence of data amount of DNN training set on positioning performance of the system.The results show that when the data amount is reduced by 80%,the average positioning error of the system is 1.75cm,and the test points with positioning error less than 5cm account for more than 97%.This has guiding.significance for practical applications,that is,according to the actual scenarios and requirements,the positioning performance and labor cost of data collection can be balanced,and the appropriate amount of data can be selected.