自适应三维鬼成像方法研究

Compared with the traditional three-dimensional (3D) imaging, 3D ghost imaging (3DGI) obtains higher resolution because of breaking diffraction limit. Moreover, with the advantages of good anti-interference and sensitivity, etc., shows the promising in the fields of military and civilian, such as remote sensing, reconnaissance and biomedical engineering. However, the performance of current 3DGI cannot meet the requirements of many applications because of its mechanism. The main challenges embody in two aspects. On the one hand, low efficiency is the main drawback because of contradiction between mass sampling and real-time. On the other, traditional 3DGI is reconstructed by the use of uniform slices in time-domain. However, low ranging accuracy of 3DGI is resulted by the aliasing among different slices under the situation that the pulsed laser echoes are broaden..Therefore, originated from the mechanism of 3DGI, we propose a novel approach combining self-adaptive space-variant resolution in space domain and self-adaptive differential pulsed laser echoes in time domain. In space domain, based on the principle on bionic vision with space-variant resolution, optimized random speckle patterns are obtained by the use of weighted multiparameter method. In time-domain, we deduce a full-wave models on objective significance. Self-adaptive slicing GI is achieved by studying the relationship between ranging information and pulsed laser echoes. Meanwhile, we propose a receiving method based on differential optical-path structure. Therefore, the accuracy of 3DGI is improved, because the effects from the pulse broaden and noise is suppressed. We develop a prototype to test the correctness and reliability of the proposed method, which paves the way for designing high performances of 3DGI for practical use.

相比较传统三维成像，三维鬼成像因能够突破衍射极限而实现更高分辨成像，同时具有抗干扰性强、灵敏度高等优势，在侦察、遥感及生物医疗等军民领域具有明确的应用前景。然而，现有三维鬼成像方法因受成像体制制约，导致其综合性能仍然难以满足应用需求。突出表现在：（1）大量采样与快速成像互为矛盾，导致高分辨率与实时性难以兼顾。（2）传统三维重构采用均匀切片，导致难以避免回波脉宽展宽引起的三维重构精度低下问题。对此，本项目从成像体制源头创新，提出空域自适应与时域自适应相结合的三维鬼成像方法。空域上，建立异构变分辨率鬼成像模型，通过多参数权重法实现随机散斑优化；时域上，建立基于目标显著性的全波形模型，通过研究波形显著性实现具有目标导向功能的非均匀片成像。同时，设计光路差分结构抑制噪声与脉宽展宽对获取切片位置的影响，从而实现高精度三维重构。通过实验验证提出方法的合理性与可靠性，为实现高性能三维鬼成像提供技术支撑。

本项目研究了人眼视网膜成像机理、空域变分辨和时域非均匀切片方法和实现技术并研制了一套自适应三维鬼成像原理样机并进行了综合测试与验证，最终实现了成像分辨率为60环×128（像素），成像速率达到125fps的成像结果，并且验证了仿人眼视网膜对数极坐标模型的旋转与尺度不变性。本项目将人眼视网膜机理和非均匀切片和传统计算鬼成像结合，改善了传统鬼成像的成像性能，为我国航空航天和军事国防等领域提供一套快速的新型光学成像系统。.本项目突破了仿人眼变分辨率成像、时域非均匀切片等关键技术。其中，仿人眼变分辨率成像技术的突破，可为复杂环境场景中难以快速获取小目标图像、感兴趣区域成像质量低等问题提供解决途径；时域非均匀切片技术，可为长纵深距离下实现有效目标的快速获取提供解决途径.可见，通过对本项目的研究，有望在远距离的小目标成像、城市作战中的非视域成像、导弹追踪目标等国防及民用场景中得到应用。同时，本项目通过引入GUP加速算法，结合鬼成像的简单结构形式，应用在显微眼科手术的多模态成像与信息融合，实现了面向生物医疗高质量与高效图像获取与感知。