Research On Target Following Technology Of Indoor Wheeled Robot Based On Vision

In order to make the mobile robot serve people better,it is necessary to solve the problem that the robot follows the target pedestrian movement to improve the robot's autonomous movement and human-robot interaction ability effectively.Robots with follow-up function can be widely used in elderly home care,smart photography,public services and other fields.This technology has broad application prospects.(1)Monocular pedestrian position calculation: A method for calculating relative position between pedestrian and robot based on the target pedestrian height as the prior scale information by monocular vision is given.The relative position is calculated according to the pinhole camera model which is used for calculating the motion control command of the robot;(2)Pedestrian tracking: A pedestrian tracking algorithm is proposed based on the pedestrian detection algorithm of convolutional neural network(SSD),get the similarity by extracting the position and color features of pedestrians and performing weighted fusion,then use the similarity to match candidate pedestrians provided by detection algorithm to realize the robust target pedestrian tracking in the image plane;(3)Robot motion control: A robot motion control method is proposed for the robot's target pedestrian follow up mission.According to the pedestrian position,the motion control command of the wheeled chassis is calculated in real time,makes the robot accelerates when the pedestrian far away,decelerates when close so that to reduces the probability of target loss as much as possible.The results of simulation experiments and actual tests of wheeled robots based on ROS platform show that the method proposed by the paper realize the wheeled robot follow the target pedestrian motion in indoor environment effectively based on monocular vision.The research production of this paper have been applied to the intelligent mobile shooting system of a radio and television station,which is used to realize the automatic tracking-shooting function of the robot.