The Research On Direction Measurement Technology Of Large Caliber Gun Barrel Based On Machine Vision

The firing data of fire control system is an important test index,therefore,in the design,calibration and application of artillery,there is a high demand for the precision of the firing data solution method,aiming at the problems of complex installation and debugging,low measurement precision and unsuitable for dynamic measurement,this paper studies the non-contact measurement method based on machine vision,the practicability and robustness of the method are proved by many experiments.Firstly,a strip marker is designed as a cooperative target,and the center of the circle is used as an image feature point to solve the pose,the motion posture of the exit tube is calculated by a certain function relation with relative parameters.In order to analyze the factors related to the accuracy of attitude measurement,the analytic expression between direct measurement and indirect measurement is established,in order to avoid the complexity of PNP function,the error characteristics of characteristic point connection are analyzed,and the error rules affecting attitude measurement accuracy are obtained.Then,a method of adaptive threshold preprocessing and feature point correction is proposed to solve the problem that the precision of target point extraction is not accurate in complex environment.In this method,the mark containing"x"-shaped feature points is fixed on the gun barrel as a target.After searching the optimal transformation parameters to adapt to illumination by grid,the initial corner coordinates are detected by corner detection algorithm,considering the geometric constraints between feature points and the distribution characteristics of Pixel gray values,the linear fitting method is used to correct the Strip marks,the intersection point of the fitting curve between the corners is obtained by using the non-linear function regression maximum likelihood as the image coordinates of the corrected center of the circle,by using the orthogonality of the rotation matrix and the least square algorithm,the corresponding external parameter rotation matrix is obtained.Finally,the effectiveness and accuracy of the system are verified in real time by simulating the gun barrel with the rocker tripod.Secondly,aiming at the problem that it is difficult to determine the physical coordinates of the feature points on the barrel surface by conventional measurement methods,a method based on multi-determined scene is proposed,by setting checkerboard in the scene,the pose parameters of the image scene are determined,and the initial three-dimensional coordinates of the space point are calculated with the image coordinates of the center of the circle obtained by manual interaction.Based on this method,a inclinometer is used to maintain the special angles of the barrel,such as 0°、15°、30°,etc.the tolerance constraints between the measured and determined attitude angles from the initial coordinates are transformed into a penalty function,and the x axis coordinates are taken as the particle dimensions,within a certain range of initial coordinates,the particle swarm optimization(PSO)algorithm is used to solve the constrained optimization problem,and the initial coordinates are optimized,the distance error between the marked position of x axis and the real position is less than 3mm,the accuracy of Y axis and Z axis is less than 1mm.Finally,the design and implementation of the body-oriented vision measurement system are studied.Firstly,the general scheme of the system is presented,which consists of a main control computer,an on-line testing device and several visual measuring terminals,and a large caliber gun barrel pointing measurement system based on C/S architecture is constructed,synchronous measurement and data processing,combined with the development of the master computer software and visual measurement terminal software,the detailed operation process is given.