Research On Theory And Technology Of Visible Light Positioning Based On Image Sensor

LED(Light-Emitting Diode)visible light positioning technology which is based on Visible Light Communication(VLC)is aimed to locate the target receiver using the position of the white LED and the electrical signal parameters carried by the optical intensity.White LED-based visible light positioning is one of the most potential positioning technologies in the future,and is characterized by high positioning accuracy,implementation simplicity and no electromagnetic interference etc.,and can be widely used in the public places such as large scale shopping malls and subways,and the places susceptible to radio frequency electromagnetic interference such as hospitals and scientific laboratories.The unique spatial separation of the Image Sensor(IS)makes it to spatially separate the signals from different white LED light sources,and to spatially separate the signal sources from the noise sources so as to detect the useful signal and effectively eliminate the noise.Furthermore,the widespread of camera or IS utilized in the mobile terminals or wearable devices makes it available to meet the practical application requirements to launch research on the IS-based LED visible light positioning.The IS-based LED visible light positioning is a multi-subject comprehensive technology integrated into LED visible light communication,computer vision and photogrammetry etc.How to determine the position or attitude of the IS receiver(for short imaging receiver or camera receiver),and to improve the positioning accuracy or attitude accuracy is still a difficult point in current research work.In summary,the research on the IS-based LED visible light positioning technology has important theoretical significance and application value.In this paper,a parameter estimation algorithm of the position and the attitude of the imaging receiver is successively presented,which will lay statistical analysis foundation for the IS-based LED visible light positioning technology.Furthermore,the theoretical limit of the estimation accuracy of the position and the attitude,i.e.the Cramér–Rao lower bounds(CRLB),is respectively derived and the determination of the CRLB will undoubtedly guide the direction for the parameters optimization such as the LED,the IS and the communication distance between the LED transmitter and the imaging receiver.On this foundation,we propose a method of I2V(Infrastructure-to-Vehicle)vehicle position estimation,and also present a V2V(Vehicle-to-Vehicle)adjacent vehicle distance and orientation estimation algorithm,all of which will provide theoretical basis for the outdoor application of the IS-based LED visible light positioning technology,and give important practice instruction significance.The main research contents and contributions are listed as follows:(1)For the observation values of the LED's imaging point being affected by white Gaussian noise,a Maximum Likelihood Estimation(MLE)algorithm of the position of the indoor imaging receiver is first proposed,which lays statistical analysis foundation for the IS-based LED visible light positioning technology.Then,the theoretical limit of the estimation accuracy of the position i.e.the CRLB is determined and the CRLB provides an absolute lower bound to which the mean square error of any unbiased estimator can attain.Finally,we make simulation experiment with realistic values of the LED and the image sensor for a typical indoor scenario,and the simulation results show that the positioning accuracy of the indoor imaging receiver depends on the vertical distance from the LED transmitter plane to the imaging receiver plane,and on the focal length of the lens,the pixel size and frame rate of the imaging receiver.Accurate positioning estimation of the indoor imaging receiver is achievable,with positioning accuracy usually in the order of millimeters.(2)In order to accurately determine the spatial state of camera or IS in the three-dimensional space,this paper proposes a joint maximum likelihood estimation algorithm for the position and the azimuth rotation angle,and realizes the attitude estimation of the indoor imaging receiver.Furthermore,the theoretical limit of the estimation accuracy of the attitude i.e.the CRLB is determined.Finally simulation experiment with realistic values of the LED and the image sensor for a typical indoor scenario is performed and the simulation results show that the estimation accuracy of the attitude of the indoor imaging receiver depends on the number of LED used,and on the focal length of the lens,the pixel size and frame rate of the imaging receiver.Accurate attitude estimation of the indoor imaging receiver is achievable,with the estimation error of the position usually in the order of centimeters,and the estimation error of the azimuth rotation angle being less than 1~0.(3)For the outdoor I2V vehicle position estimation,the Maximum Likelihood Estimation(MLE)and Least Square Estimation(LSE)algorithms are successively presented to localizate a vehicle using the signal from the LED array of the traffic light in a crossing and the image of the LED in the imaging receiver mounted on the vehicle.Then the CRLB of the positioning estimation accuracy of the vehicle is determined and compared with the Mean Square Position Error(MSPE)of the MLE and LSE algorithms.Finally,we make simulation experiment with realistic values of the LED array of the traffic light and the image sensor for a typical outdoor I2V scenario,and the simulation results show that with the communication distance being constant between the LED array transmitter and the imaging receiver mounted on a vehicle,the positioning estimation accuracy of the outdoor I2V vehicle depends on the number of LED used,and on the focal length of the lens,the pixel size and frame rate of the imaging receiver mounted on the vehicle.Accurate position estimation of the outdoor I2V vehicle is achievable,with the estimation error of the vehicle position generally less than 0.1 m at a communication distance of 30 m.(4)For the outdoor V2V adjacent vehicle distance and orientation estimation,a novel method of distance estimation is first presented to determine the V2V adjacent vehicle distance using the image of the LED of the vehicle ahead in the imaging receiver mounted on the following vehicle.And the orientation rotation angle between the adjacent vehicles is then derived through a maximum likelihood attitude estimation algorithm.Finally the numerical simulation results show that distance estimation accuracy depends on the accuracy between the imaging center of the LED,and the estimation accuracy of the azimuth rotation angle is affected by the adjacent vehicle distance,with the estimation error of the azimuth rotation angle generally being less than 1~0 at a distance of 35 m between the adjacent vehicles.