Research And Implementation Of Localization Method Based On LED And Event Camera

Event Camera is a new kind of vision sensor inspired by biology,which has attracted more and more attention in recent years.It is different from the traditional camera that captures the image frame at a fixed rate,but outputs the event stream of information about the change of light intensity at each pixel in an asynchronous form.This perceptual form gives it great advantages over traditional cameras such as low latency(20us),high temporal resolution,and high dynamic range(120d B).Based on the above advantages of event camera,using event camera to replace traditional CMOS camera for positioning and tracking has become a new research direction in the field of visible image positioning.This paper proposes a visible light terminal side positioning and tracking method based on event data.The main research work is as follows:Firstly,according to the different Visible Light Positioning(VLP)receivers,the research status of PD-based VLP technology and image sense-based VLP technology at home and abroad are introduced respectively.Several common visible light signal localization algorithms based on PD photodetector and image sensor are introduced,and the complexity of the two types of localization algorithms and their applicable positioning scenes are analyzed.Among the visible light indoor localization methods based on PD,TDOA algorithm has very strict requirements on the estimation accuracy of signal arrival time delay,with high positioning accuracy.The positioning algorithm based on image sensor has large data transmission,slow speed,limited dynamic range and high positioning accuracy.Then,aiming at the problems that the indoor positioning algorithm based on image sensor has limited dynamic range,high delay and is easily affected by various lighting and environmental factors,a visible light image positioning algorithm based on asynchronous event data is proposed,which effectively reduces the positioning delay.Moreover,compared with CMOS cameras,event cameras are not affected by motion blur caused by high-speed motion.It is still applicable in high speed motion and high dynamic range scenarios.In a 700 mm x 700 mm x 1000 mm positioning scenario,the average positioning error can be less than 3cm,and the average positioning delay can be less than 10 milliseconds.In scenarios with different positioning positions and terminal velocities,the average positioning error is stable in the order of centimeters,and does not increase with the increase of terminal velocities.At the same time,considering the possibility of large outlier errors in the actual scene of LED-ID,a positioning optimization strategy based on median selection was proposed to estimate the optimal positioning point of the camera.The proposed method can further improve the accuracy and stability of positioning,and the positioning error is stable within 2cm.Finally,the Particle Filter(PF)algorithm is combined with the visible light location algorithm based on event data to realize the trajectory tracking of the moving target.Each LED-ID pixel coordinate detected by the camera after dedistortion is taken as the observation value of the PF algorithm.Based on the observation value at each sampling moment,Update particle weight to achieve the optimal estimation of terminal position.In this paper,it is verified that in the initial state,the tracking particle velocity is accurate,the position is inaccurate and the position is accurate,the velocity is inaccurate,and the algorithm finally reaches convergence.Furthermore,in order to improve the real-time performance and robustness of positioning and tracking in complex tracking scenarios,an improved PF algorithm is proposed by embedding Mean Shift(MS)algorithm into particle filter.Through experimental simulation analysis,the proposed algorithm not only effectively reduces the computational load,but also effectively improves the tracking performance of mobile terminals.It has strong practical significance.