Imaging through atmospheric turbulence is a topic with a long history and grand challenges still exist in the remote sensing and astro observation fields. In this letter, we try to propose a simple scheme to improve the resolution of imaging through turbulence based on the computational ghost imaging (CGI) and computational ghost diffraction (CGD) setup via the laser beam shaping techniques. A unified theory of CGI and CGD through turbulence with the multi-Gaussian shaped incoherent source is developed, and numerical examples are given to see clearly the effects of the system parameters to CGI and CGD. Our results show that the atmospheric effect to the CGI and CGD system is closely related to the propagation distance between the source and the object. In addition, by properly increasing the beam order of the multi-Gaussian source, we can improve the resolution of CGI and CGD through turbulence relative to the commonly used Gaussian source. Therefore our results may find applications in remote sensing and astro observation.

We experimentally demonstrate that ghost imaging (GI) quality can be significantly improved by modulating the incoherent Gaussian sources with hyperbolic cosine functions. In our experiments, a source with desired cosh-Gaussian intensity distribution can be generated by a programmable spatial light modulator, and is used in the GI setup to obtain high quality ghost images. The influences of the hyperbolic cosine parameter on the point-spread function and the GI quality have been investigated experimentally and interpreted theoretically. Our experimental results show that the resolution of ghost images can be obviously enhanced with cosh-Gaussian modulated incoherent sources. The kind of source-shaping technique proposed in this paper is a high-efficiency method to improve the GI quality and may be very useful for the GI applications.

We develop an adaptive differential correspondence imaging (CI) method using a sorting technique. Different from the conventional CI schemes, the bucket detector signals (BDS) are first processed by a differential technique, and then sorted in a descending (or ascending) order. Subsequently, according to the front and last several frames of the sorted BDS, the positive and negative subsets (PNS) are created by selecting the relative frames from the reference detector signals. Finally, the object image is recovered from the PNS. Besides, an adaptive method based on two-step iteration is designed to select the optimum number of frames. To verify the proposed method, a single-detector computational ghost imaging (GI) setup is constructed. We experimentally and numerically compare the performance of the proposed method with different GI algorithms. The results show that our method can improve the reconstruction quality and reduce the computation cost by using fewer measurement data.

Computational ghost diffraction (CGD) with a higher-order cosh-Gaussian modulated incoherent source is investigated theoretically. The corresponding numerical simulations are given to see clearly the effects of the parameters of the higher-order cosh-Gaussian source on the imaging quality. Our results show that the resolution of the CGD patterns can be significantly improved by properly varying the source parameters. In addition, we numerically study the effect of the propagation distances in the CGD system and explore the CGD applicability in coherent diffraction imaging. These results may be helpful for implementation of high-resolution x-ray diffraction.

Transportation panoptic perception (TPP) is a fundamental capability for both on-board and roadside monitoring systems. In this paper, we propose an end-to-end lightweight multitask model, MT-TPPNet, which jointly performs three tasks: object detection, drivable area segmentation, and lane line segmentation. To accommodate task differences while sharing a common backbone, we introduce the Asymmetric Projection with Expanded-value (APEX) mechanism, which integrates attention mechanisms with different biases to enhance performance across various tasks. We further propose the Selective Channel–Spatial Coupling (${\mathrm{SC}}^2$) mechanism, which injects complementary frequency-band information into the channel-spatial coupled features. Additionally, by using a unified loss function to simultaneously handle detection and segmentation tasks, we eliminate the need for task-specific customizations, improving both training stability and deployment flexibility. Extensive experiments on self-collected field data and public benchmarks from roadway and railway scenarios demonstrate that MT-TPPNet consistently outperforms strong baselines in terms of mAP, mIoU, and FPS. In particular, MT-TPPNet achieves a mAP50 of 83.2% for traffic object detection, a mIoU of 91.6% for drivable-area segmentation, and an IoU of 28.9% for lane-line segmentation, demonstrating the effectiveness of the proposed approach.