Research On Visible Light Multi-User 3D Localization Algorithm Based On Monocular Vision

With the development of the Internet of Things and artificial intelligence,the demand for positioning technology has increased dramatically,especially indoor positioning technology.Due to the interference of buildings,the developed Global Positioning System(GPS)has large indoor errors.Compared with indoor positioning technologies such as WiFi,Ultra Wideband(UWB),Zigbee and Bluetooth,Visible Light Positioning(VLP)technology has advantages in cost,accuracy and security.Visible light positioning uses low-cost,high-efficiency,low-carbon,and environmentally friendly Light Emitting Diode(LED)as the signal source,and uses currently installed surveillance cameras as the receiving end.This paper focuses on the monocular vision multi-user 3D positioning algorithm in the visible light band.The main work is as follows:(1)Design and implement an encoding and decoding scheme combining UPSOOK imaging recognition and color recognition.The experimental results show that the scheme can be used in At the same time of communication,it can realize multi-user identification with high accuracy and can resist the problem of color identification in different environments.(2)When the surveillance camera locates the user,the sending module tilts,which affects the positioning accuracy.Aiming at this problem,a high-precision anti-tilt and motion blur positioning algorithm based on neural network is proposed.The experimental results show that the algorithm can accept the inclination of the XY axis within 15 degrees and the 360 degree rotation of the Z axis.Therefore,it has good resistance to the rotation and tilt problem of the positioning target.When the target is at different rotation angles and under different tilt angles,the algorithm can ensure that the mean positioning error(MPE)of positioning is 0.87cm.The Z-axis rotation in conventional motion positioning and the motion blur caused by motion can well resist the positioning error caused by rotation and motion blur through data preprocessing,which improves the value of practical application scenarios.(3)A multi-user positioning system scheme based on monocular vision is designed and implemented,and an experimental system is built.The system adopts simple image processing algorithm and the anti-tilt pose estimation algorithm based on neural network proposed in this paper.The experimental results show that it has a good positioning effect for multiple users,and the positioning MPE is within 1cm.