Research And Realization Of Vision Guidance Method For Automatic Missile Loading

As an important long-range strike method in modern warfare and one of the carriers of strategic threat weapons,missiles are an important part of our country’s national defense force.Missile loading is very important in the manufacturing process of missiles.Its task is to assemble the missile into the cartridge.How to accurately dock the cartridge with the missile track frame is a key issue.At present,the domestic missile loading work is mainly to adjust the simple tooling mechanism manually,and the man pushes the truck for the final docking.This has extremely high requirements for the technical level of the workers,especially for the loading of large missiles,which takes a long time.With the development of computer vision technology,more and more computer vision algorithms are used in industrial production.Therefore,if an automated missile loading method can be developed,it will greatly improve the production efficiency of missile assembly.This is of great significance to the development of our country’s national defense industry.This topic researches and realizes the relative pose measurement method between the cartridge and the missile track frame.This paper proposes a monocular vision relative pose measurement method based on a fixed target,which uses less hardware resources to perform high-precision real-time relative pose measurement.The research content mainly includes:1.Target design.First,a ring target is designed based on the hardware characteristics of the application platform and the relative measurement performance requirements.In order to avoid the difficult problem of detecting the precise position of the edge of the ring under complex lighting conditions,a series of corner features are designed on the edge of the ring as the detection target.These corners can be used to fit the inner and outer ellipses with high accuracy.2.The method of extracting and tracking feature points of the target plane.First,the Hough gradient method is used to coarsely locate the center of the ring;then,the subpixel position of the corner feature is detected and the RANSAC method is used to screen the feature points.Design the target tracking method,compare the template matching method and the L-K optical flow tracking method based on the image pyramid in the actual application of the missile track frame translation process,and finally choose the optical flow tracking method to track the target.3.A ring modeling method based on the invariance of the intersection ratio of the four collinear points is used to solve the problem of the relative pose measurement of the ring target.First,use the detected corner point features to fit the inner and outer ellipse;then,according to the invariant nature of the four collinear points,calculate the projection point position of the target center in the image;finally,using the position of the projection point of the target center,the spatial position of the target center and the edge of the ring is calculated using the ring modeling method,and then the relative pose is calculated,and the Kalman filter is performed on the measurement result to improve the stability of the measurement result.In order to verify the feasibility of the method and meet the requirements of technical indicators such as real-time,stability and accuracy,this paper designs and completes simulation experiments and physical experiments.The experimental results show that the system achieves millimeter-level position accuracy within the working range of 5-10 cm,the processing time is less than 100 ms,and it is adaptable to changes in illumination,and the measurement results are stable,which meets the needs of practical applications.