基于粒子群优化BP神经网络的激光扫描投影系统畸变预测方法

Distortion Prediction Method of Laser Scanning Projection System Based on PSO-BP Neural Network

为了精准、高效地预测和校正激光扫描投影系统的畸变误差,研究了基于粒子群优化BP神经网络的畸变预测方法。建立了BP神经网络结构,并融合粒子群优化算法对BP神经网络的权值和阈值进行优化,得出基于粒子群优化BP神经网络的激光扫描投影系统投影畸变预测模型。选取距激光扫描投影仪器两米的待投影面上的理论坐标点及各点相应畸变值Δx作为粒子群优化BP神经网络的训练数据集,将待投影面上实际投影位置坐标代入训练好的粒子群优化BP神经网络进行预测得到预测畸变值输出,并与实际畸变值对比,最后,引入Elman神经网络预测模型的预测结果与所研究预测方法进行对比。结果表明:在±30°的全视场扫描投影范围内粒子群优化BP神经网络预测模型的均方根误差为0.0176 mm,解算时间仅需22.4 s,相较于Elman神经网络效率提升78.33%,预测精度及时间明显优于Elman神经网络,可以有效预测激光扫描投影系统的畸变误差。

Laser scanning projection technology can accurately project the patterns of workpiece,text about the processing and other information on the target location based on the CAD model,so the technology is widely used in the advanced manufacturing and intelligent assembly.However,there are theoretical projection distortion errors in the laser scanning projection system,and the distortion errors seriously affects the accuracy of the shape and position of the projected patterns.In order to ensure the accuracy of the projected patterns,it is necessary to predict and correct the distortion errors of the laser scanning projection system accurately and efficiently.Nevertheless,the distortion correction methods for 2D galvo scanner projection system are only commonly reported in the research such as LiDAR and laser marking and other technologies.For the distortion prediction and correction methods of 2D galvo scanner projection system,which are studied in this paper,there is rarely reported both domestically and internationally.In view of this,the particle swarm optimisation BP neural network approach is used in this study for the prediction and correction of distortions in laser scanning projection graphics.In recent years,Elman neural networks have been used in some related studies to correct the distortion error of 2D galvo scanner.Therefore,the Elman neural network algorithm by the same training and test process is adopted in the comparative experiment,and then the two sets of prediction accuracy are compared.In this study,the particle swarm optimization algorithm is studied to optimize the weights and thresholds of BP neural network,and the problems of easily falling into local extremes and overfitting are solved.At the same time,particle swarm optimization BP neural network is also used to predict the distortion errors of the projection pattern in the laser scanning projection system,thus the prediction accuracy of the studied PSO-BP neural network algorithm can be proved.In this study,it is necessary to determine the training data set and test data set of the neural network,obtain the coordinate values with distortion errors which calculated by the coordinate transformation formula,and then calculate the corresponding distortion errorΔx as the training data set.The binocular vision measurement system is used to obtain the actual coordinate values with distortion errors,and the corresponding distortion errorΔx is calculated as the test data set.Then the number of hidden layers,N,is changed by many times,multiple root mean square error values of prediction accuracy are obtained by the studied method according to different number of hidden layers.When the root mean square error is the smallest,the selected value of N is determined as the number of hidden layers of the particle swarm optimization BP neural network.The neural network is trained by the training data set,and in order to verify the generalization ability of the trained neural network,the test data set is used to test the neural network to avoid the overfitting problem.The particle swarm optimised BP neural network model established in this paper is trained by both training and test datasets,and the Root Mean Square Error(RMSE)of the prediction accuracy can reach 0.0176 mm,and the calculated time is only 22.4 s.Meanwhile,in the comparison experiments with Elman neural network,the RMSE of Elman's algorithm is 0.6826 mm.The particle swarm optimization BP neural network algorithm improves the prediction accuracy by approximately 97.419%which compared to the Elman neural network algorithm.In the paper,a distortion errors prediction method for laser scanning projection system based on particle swarm optimization BP neural network model is proposed,through the validation experiments with the Elman neural network,the results show that the root-mean-square error of the PSO-BP prediction model is 0.0176 mm,and the calculated time is only 22.4 s,but the root-mean-square error of the Elman algorithm is 0.6826 mm.The particle swarm optimization BP neural network algorithm improves the prediction accuracy by about 97.419%compared with the Elman neural network algorithm.Compared with the traditional Elman neural network algorithm,it can predict the distortion errors of the laser scanning projection system more accurately,and can be applied to the developed laser scanning projection system to solve the problem of accurate correction of the distortion errors.The studied method can also significantly improve the shape accuracy and position accuracy of large-scale projection,can enable the developed laser scanning projection system to perform more precise digital assembly and intelligent positioning operations.