Research on fast prediction method for far-field spot characteristic parameters in laser atmospheric propagation

The rapid evaluation of far-field spot characteristic parameters is essential for modern laser engineering applications. This study establishes a multi-layer phase screen numerical simulation model, a support vector machine (SVM) prediction model, and a random forest (RF) prediction model based on experimental data from laser propagation in the field. The performance of these three fast evaluation methods is compared and analyzed. The results indicate that the prediction models constructed using the random forest and support vector machine algorithms outperform the phase screen numerical simulation model. Specifically, the random forest prediction model achieves an average relative error of no more than 10% and a modeling time of less than 11 seconds, which is superior to the support vector machine prediction model. These findings provide valuable technical support for the rapid prediction of far-field spot characteristic parameters in laser atmospheric propagation.