Research On Localization Method Of Epidemic Prevention Mobile Robot Based On Distributed Surveillance Camera

With the development of network information technology,the surveillance camera market is becoming larger and larger,with various types of IP cameras all over the streets,not only outdoors,many enterprises,shopping malls,and other indoor scenes are also equipped with network surveillance cameras.The outbreak of the epidemic in 2020 has led to the installation of surveillance cameras in isolation wards of many hospitals to better observe the conditions of patients.At the same time,as mobile robot-related technologies become more mature and the market gradually expands,a variety of mobile robots have emerged,from sweeping robots to food delivery robots to epidemic prevention and disinfection robots.Using surveillance cameras distributed indoors to locate the epidemic prevention mobile robot is not only accurate,suitable for multi-robot positioning,but also saves hardware costs.However,surveillance cameras on the market generally have excessive image transmission delay.If the traditional distributed camera positioning system is directly used to locate the epidemic prevention mobile robot,it will inevitably cause a serious positioning lag effect,resulting in the inability to effectively control the epidemic prevention movement.robot.On the other hand,robot vision capture technology based on traditional marker generally has problems such as too short recognizable distance,poor anti-interference ability,small number of codes that can be coded,and too large footprint.The main research topic of this paper revolves around the method of using distributed surveillance cameras for global positioning of epidemic prevention mobile robots,and solves two key technical problems.One is the positioning lag caused by the delay of the surveillance camera,and the other is the problem of the camera’s visual capture of the robot.Aiming at the delay problem of surveillance cameras,this paper proposes a robot positioning algorithm based on multi-camera and IMU delay compensation.Firstly,the average delay time of each camera is recorded;the key points of the robot are recognized on the images of each camera,and the time stamp is stored in the buffer pool.Then take the camera with the longest delay time as the standard for simultaneous sampling.Thirdly,use the camera model to obtain the measured pose of the robot through SVD decomposition.Fourth,use Kalman filter to estimate the lagging pose.Then compensate the lagging posture information through the IMU.Experimental results show that when the camera has high latency,the dynamic positioning error of the algorithm is reduced by 85.6% compared with the traditional algorithm,which significantly improves the effectiveness of the positioning system.Aiming at the problem of robot vision capture,this paper designs a dynamic LED vision marker and proposes its tracking and positioning algorithm,which can locate and distinguish all robots equipped with LED visual marker in the image.The dynamic LED visual marker uses colored LEDs as the carrier,and transmits information through the sequential arrangement of three states of red,green and blue.At the same time,a coding rule based on a trigeminal tree is designed.The tracking and positioning algorithm adopts the dual-thread design of the tracking thread and the detection thread.The tracking thread uses the Kalman filter algorithm to track the LED marker in the image while completing the encoding verification.The detection thread locates the LED and reads the encoded information by detecting the start signal.The thread design effectively reduces the amount of calculation,but also has the characteristics of relocation and fast response.The experimental results show that the LED dynamic marker is used to locate the external camera of the robot,and the positioning accuracy is 1.6 cm,which fully confirms the effectiveness and accuracy of the beacon and its tracking algorithm.