Portable and Adaptive Data Analysis Workflows for Real-Time 3D Vision (B02)

在许多自然科学中，显微镜是发现以前从未被相机看到或捕捉到的东西的关键。在高端显微镜中，分辨率接近探测辐射的波长，从而导致以部分相干、潜在单向或甚至偏振照明记录的图像受到各种衍射效应的困扰。处理这种数据的重建算法是计算和/或存储密集型的，并且通常是供应商特定的黑盒算法，其不可能移植并且难以修改或维护。这个跨部门子项目的目的是开发一个工具箱和平台，用于灵活创建与供应商无关的数据aws，以便对显微图像进行近实时分析。它解决了显微镜之间的便携性，分析管道的可靠性，以及适应不同的测量类型和可用资源。将与B5在资源需求估算方面以及与B4在网络流量优化方面建立重要合作。该分项目将由视频流在线操作和分析专家Alfret教授和电子显微镜和光学显微镜新技术开发专家Koch教授执行。

In many Natural Sciences, microscopy is key to discover things that have never been seen or captured by cameras before. In high-end microscopy, resolution approaches the wavelength of the probing radiation, leading to images that are recorded with partially coherent, potentially unidirectional, or even polarized illumination plagued by various diffraction effects. The reconstruction algorithms processing such data are compute- and/or memory-intensive and are typically vendor-specific black-box algorithms, which are impossible to port and hard to modify or maintain. The aim of this interdiscipli-nary subproject is to develop a toolbox and platform for the flexible creation of vendor-agnostic DAWs for the near-real time analysis of microscopic images. It addresses portability across microscopes, dependability of analysis pipelines, and adaptation to different measurement types and available re-sources. Important cooperations will be established with B5 on estimating resource requirements and with B4 on optimizing network traffic. The subproject will be carried out by Prof. Eisert, an expert in online manipulation and analysis of video streams, and Prof. Koch, an expert in developing novel techniques in electron microscopy and optical microscopy.