Research on the performance enhancement of rotational Doppler effect based on a joint channel-spatial attention joint optimization algorithm

Abstract Atmospheric turbulence not only induces severe wavefront distortion and orbital angular momentum mode dispersion in vortex beams, but also poses a significant challenge to the accurate measurement of rotational velocity. Existing adaptive optics restoration algorithms suffer from issues such as a tendency to converge to local optima and limited restoration quality of the optical field. Conventional convolutional neural network models are inadequate for effectively capturing the complex multi-scale distortion features of distorted wavefronts. In response to the distribution characteristics of turbulence-induced perturbations across channel and spatial dimensions, this study proposes a joint channel-space attention optimization algorithm. This approach enhances the perception and extraction capabilities for high-order perturbations and localized severe distortion regions from both channel and spatial dimensions. Compared to the baseline model and single-attention enhancement models, the proposed method demonstrates superior predictive and generalization performance. Even under strong turbulence conditions with D / r 0 = 7, the mean squared error, root mean square, and high fluctuation ratio are only 58.51%, 61.33%, and 60.21% of those of the baseline model, respectively. When D / r 0 = 10, the P IF , SMSR IF , and σ IF reach 31.85, 28.65, and 0.3, respectively. Compared to channel attention, the spatial attention mechanism exhibits better performance in optical field restoration and spectral improvement. The joint optimization of both mechanisms further mitigates the degradation in prediction performance caused by increasing turbulence intensity and enhances velocity measurement accuracy.