稀疏贝叶斯学习实现窄带射频多态衰落综合成像的关键技术研究

This project is to address the key issues in joint multi-modal fading tomographic imaging with sparse Bayesian learning, which enables the exploration of a new imaging paradigm for ghost suppression. In particular, firstly, the dual imaging model will be studied under the framework of sparse Bayesian learning for facilitating multi-modal imaging tasks. Then, how to jointly perform the tasks of configuring project matrices and reconstructing fading images will be investigated for online updating the project matrices adaptively. Finally, the cooperative mechanism between the parameter estimation tasks of signal model and measurement model will be explored to form the integration paradigm of performing the tasks of objective function design and image reconstruction. Towards on simultaneously reconstructing fading images and suppressing ghost interference, the key to multi-modal fading tomographic imaging with sparse Bayesian learning, is to address the challenges arising from the different objectives between the individual imaging task, so as to form a compatible learning mechanisms and implementation schemes, which is the main aim of this project. The novelty and significance of this project is attributed to the efforts that the multi-path fading is treated as informative cues and will be transferred as the potential information for eliminating ghosts. ..The output of the project will be a novel yet practical solution for the suppressing ghosts in RF imaging tasks, the results of the project will have definitively theoretical and practical significances in various applications.

研究窄带射频传感网络实现多态衰落综合层析成像的稀疏贝叶斯学习方法，探索对抗伪影干扰的新模式。特别地，研究对偶状态影像重构的稀疏贝叶斯学习框架，形成兼容多态衰落影像重构任务的信号表示模型；探究投影矩阵配置与影像重构一体化的学习架构，形成多态衰落影像投影矩阵的在线适应配置模式；探索信号模型和测量模型参数估计同步的协作机制，形成目标函数设计与影像重构的协同学习方法。面向阴影衰落影像重构和对抗伪影双重任务，稀疏贝叶斯学习实现多态衰落影像联合重构的关键在于，解决不同模态成像任务之间差异性带来的新问题，形成与之匹配的计算机制和实现方法，这是项目的预期目标。将多径衰落接纳为对抗伪影的线索，并加以转化利用，实现对抗伪影模式的新突破，这是项目成果的最终体现和方法创新的重要标志。.此方面的研究，既是窄带射频成像技术完善和发展的要求，也是其在复杂环境应用拓展的需要，有重要的学术意义和广泛的应用前景。

多径衰落诱导的影像失真已成为射频层析成像自身难以克服的技术难题，严重制约其在真实复杂应用场景中的广泛适用性。项目从模型理论和实现方法两个层面探索多模态射频层析成像所需的稀疏贝叶斯压缩传感计算机制，形成对抗影像失真的多态衰落综合成像实现策略。具体取得三方面的研究成果。首先，发展了对偶状态影像重构的稀疏贝叶斯学习框架，形成了兼容多态衰落影像重构任务的表示模型。其次，研究了投影矩阵配置与影像重构一体化的学习模式，提出了多态衰落影像投影矩阵的在线适应配置系统化解决方案。最后，建立了信号模型和测量模型参数估计同步的协作机制，形成了目标函数设计与影像重构的协同学习范式。面向阴影衰落影像重构和对抗伪影双重任务，项目将稀疏贝叶斯学习实现多态衰落影像联合重构的作为关键突破口，解决不同模态成像任务之间差异性带来的难题，形成了与之匹配的计算机制和实现方法，实现了项目的预期目标。项目成果的重要意义体现在，将多径衰落接纳为对抗伪影的线索，并加以转化利用，实现了对抗伪影模式的新突破，这也是项目创新的重要标志。

内容获取失败，请点击重试