Laser Positioning Projection Of Three-dimensional Curves By Galvanometer Scanning System Based On Machine Vision

The technology of three dimensional(3D)laser positioning projection can provide reference and guidance for workers to operate efficiently and accurately in a specific location,so it is widely used in aircraft parts assembly,composite material layering,aircraft skin hole positioning and aviation manufacturing fields.Combining the visual measurement with laser galvanometer scanning,this paper proposed the overall solutions of the 3D positioning projection under the visual AIDS,including the calibration of the laser galvanometer,the rapid on-site recalibration of the laser galvanometer,the shortcut to marking 3D target curves on curved surface and the implementation of the laser galvanometer projection prototype system,etc.The main points are as follows:1.A new calibration method for physical model parameters of laser galvanometer scanning system is proposed.Based on the analysis of the main differences between the real galvanometer scanning system and the ideal physical model,a 5-parameter physical model of laser galvanometer is given.On the one hand,the general relation between the deflection angle of the axis of one galvanometric scanner and the angular separation of the corresponding two outgoing laser beams is derived.On this basis,the deflection angle estimation algorithm based on the cone surface composed by the direction vector of all the outgoing laser beams is proposed to make the angle estimation more accurate.On the other hand,through mathematical modeling of the 5-parameter physical model,the separability of the parameters that affect the direction of the outgoing laser beam and the position of the outgoing laser beam in the optimization process is found.In order to avoid the problem of low stability of calibration results caused by the overall optimization of all parameters in the existing optimization model,the optimization models of the parameters of the direction and the position of the outgoing laser beam are proposed respectively.Through the simulation calibration experiment of galvanometer system and the actual projection positioning experiment of the centers of the plane target circles and the space closed curves,the calibration accuracy and stability of the algorithm are verified.2.A general data-driven calibration of laser galvanometer scanning system and its solution algorithm are proposed.With the help of machine learning strategy,the Single Hidden Layer Feedforward Neural Network(SLFN)is established to represent the actual laser galvanometer scanning system,in which the galvanometer control digital signal is the input of the network,and the space vector of outgoing laser beam corresponding to the control signal is the output of the network.Using the Extreme Learning Machine(ELM)algorithm,all parameters of the neural network can be obtained quickly by solving a set of overdetgermined linear equations.The influence of the number of neurons in the hidden layer on the regression accuracy is analyzed quantitatively.The results show that the selection of the number of neurons can guarantee the generalization performance of the network in a large range.Because the output of the calibration network model is the space vector of the outgoing laser beam,the training time of the network is shorter,and it can be used to solve the problems of laser triangulation and laser location projection.The experimental results of calibration experiment,target location projection experiment and 3D surface point cloud measurement show that the proposed calibration method of data-driven galvanometer has good generality and high calibration accuracy.3.A rapid on-site recalibration of laser galvanometer scanning system based on binocular stereo vision is proposed.Using this recalibration method,a laser galvanometer scanning system which is precisely calibrated the matching relationship between the control digital signal and the corresponding space vector of outgoing laser beam can be quickly combined with a binocular stereo vision system for field application based on the galvanometer scanning system.In the whole process of recalibration sampling,only two images of laser spot array need to be taken with the help of binocular system.A specially designed linear solution algorithm makes the whole recalibration method very stable and efficient.The superiority of the method is verified by the field recalibration experiment and the target location projection experiment.4.The influence of the incidence angle of laser projection on the effect of laser location projection is analyzed,and the optimal solution model of the projection pose planning problem of galvanometer system based on the laser projection angle and the effective projection area is given.On the basis,an algorithm of the projection pose estimation in accordance with the optimal projection angle state is proposed by combining with the vision measurement technology.Using the physical model of the galvanometer system proposed in the paper,the simulation experiment of the galvanometer projection pose planning is carried out.The simulation experiment results show that the incident angles of all the target projection points on the surface of the projected object are reduced in varying degrees after the projection pose planning,thus improving the overall projection effect of target curve in the process of laser positioning projection.5.A shortcut of laser positioning projection for 3D target curves on smooth free-form surface is proposed.Without a complex calibration process of galvanometer scanning system,this method directly establishes the matching relationship between the 3D coordinates of the laser spots on the surface to be projected and the digital control signals of galvanometer scanning system.The Single Hidden Layer Feedforward Neural Network(SLFN)is used to describe the matching relationship.With the linear solution mechanism of the Extreme Learning Machine(ELM)algorithm,the matching relationship can be solved quickly.Through this matching relationship,any 3D curves on the surface to be projected can be projected quickly by the galvanometer scanning system.The experimental results of laser projection of multiple curves show that the proposed method has good positioning projection effect,and the error of positioning is less than 0.5mm.Based on the in-depth study of the above key algorithms,a prototype system of laser positioning projection for 3D contour curves is developed in this paper.Using this prototype system,several experiment samples of positioning projection are carried out on different physical surfaces.Through the projection effect observed in the field and the positioning accuracy of the discrete points on the target curves,the positioning projection performance of the system is verified.