Research On State Perception Of Space Cell Robot Based On Vision

As a new type of space operation device,space cell robot can meet the needs of different scales and different kinds of on-orbit tasks through demand-oriented selfreconfiguration,and can replace the corresponding cell modules in the event of failure,so that the whole system can be restored to normal.Work,greatly reducing the cost of launch and maintenance,and enhancing the adaptability and reliability of the system.In order to meet the space-oriented multi-tasking requirements of space cell robots,the unit cell module must have certain intelligence and autonomy to realize a series of functions such as path planning,intelligent obstacle avoidance,and autonomous connection.Therefore,the space cell robot unit cell module Perceptual ability puts high demands on it.This paper designs a space cell robot cell module based on monocular vision,which integrates control,sensing,execution,communication,energy and other units.It is driven by differential control and uses the active and passive connection surfaces to achieve intercellular locking.Connection;based on the imaging characteristics of monocular camera,the corresponding mathematical model is established,and the extraction and description of the feature points are studied.According to the design characteristics of space cell robot unit cell,a camera independent calibration method based on finite state machine is designed.The calibration of the monocular camera is completed and the distortion distortion of the image is corrected using the calibration result.According to the characteristics and types of the target objects required by the space cell robot,a spatial cellular robot target image database with tag information is established.The image data with tag information is trained based on the Faster R-CNN network to realize the classification of different target objects.And detection;according to the position parameter of the target in the image and the actual size of the target,the relative distance and rotation angle between the space cell robot and the target are obtained.For the feature point matching problem in the dynamic environment,a dynamic class target region screening method is designed.The dynamic class target pixel information in the real-time image background is removed,and the accuracy of the feature point matching is improved;the motion state of the camera itself is calculated according to the geometric relationship between the camera and the matching feature point pair,and the motion state is performed by the external state information.Correction to achieve vision-based spatial cellular robot state perception.By constructing the space cell robot physics simulation platform and prototype,the feasibility and accuracy of the vision-based spatial cell robot state sensing method were verified in the Gazebo physics simulation environment and prototype experiment,and the dynamic environment and illumination noise interference experiment were designed.The robustness of the method is verified.