Research On Target Pose Measurement With Monocular Camera

Visual measurement is widely used in fields such as national defense,military,aerospace,and industrial production because of its advantages of non-contact,flexible use,economy,and rich information.Monocular vision measurement is most widely used because of its large field of view,easy operation,and simple structure.The measurement based on the feature points of the object becomes the research focus because it does not require a specific target.This subject comes from an actual engineering project.The equipment of this project works in a special environment and needs to be transferred between different stations.Therefore,a set of cable automatic docking device system is needed to realize the connection and disconnection of the power supply,signal and air source connector of the equipment.The adaptive mechanism of the automatic cable docking device has an error adaptive adjustment function to avoid the wear of the connector due to the error.The error adjustment of the adaptive mechanism requires the pose parameters before docking as its control input parameters.According to the needs of the project,a monocular target pose measurement program is adopted.The measurement target is three inner holes,and the characteristic is the fitting circle center of the inner hole edge.First segment measurement targets and extract edges,then use Zernike moments for sub-pixel edge refinement,then fit feature points,and finally use P3P(Perspective-3-Point)for pose calculation.The main research contents of this paper are as follows:Aiming at the difficulty of a single method to accurately segment measurement targets with both global and local information in a complex environment,the paper proposes an image segmentation method based on clustering and Otsu threshold fusion.Firstly,the image is clustered according to the gray value based on the image global information.Then based on the local information,the threshold segmentation between each two categories is performed.And the threshold segmented image is combined into a binary image by OR operation.Finally,the segmented image of the measurement target can be obtained by reversing the image and retaining the connected domain with the largest area.Through the experimental comparison of various image segmentation methods,the effectiveness and robustness of the proposed fusion method are verified.To extract the edges of non-cooperative targets,Zernike moments are used for sub-pixel positioning,and adaptive thresholds based on edge information are used to further improve the accuracy of target edges.The steps to solve the pose problem using three feature points are as follows.Firstly,the coordinates of the feature points in the camera coordinate system are derived from the pixel coordinates of the corresponding projection points.Then,the pose relationship between the camera coordinate system and the world coordinate system is solved by the absolute orientation problem to realize the pose measurement.Derive the measurement error model and analyze the influence of various factors on the measurement accuracy.Analyze and verify the measurement error model and the vision measurement system through simulation experiments.There is no precise instrument for pose measurement as the standard of the accuracy of the vision measurement system in this article.In the real experiment process,the actuator of the cable automatic docking device is used for indirect measurement.Take the moving distance of the actuator between two positions as the standard.It is used to measure the accuracy of the distance between two position points measured by the vision measurement system in this paper,thereby indirectly verifying the measurement accuracy of the system.By analyzing the measurement accuracy under different exposures,the robustness of the vision measurement system in this paper is verified.The experimental results show that the visual measurement system in this paper is robust and accurate.The maximum position deviation is 0.12 mm.The measurement accuracy of the attitude angle in the vertical optical axis direction can reach 0.02°,and the measurement accuracy of the other two attitude angles can reach 0.07° and 0.08°.