Data fusion of areal measured microtopographies with plane of reference using optical sensors

该项目的总体目标是使用图像处理显微镜对技术表面进行自动化重叠测量，并随后进行融合，以创建具有更大横向尺寸和最小化测量不确定性（自动拼接）的高分辨率表面数据集。研究了利用自适应光照降低单次测量不确定度的方法，以及通过数据融合使拼接不确定度（配准误差、表面偏差）最小化的方法。采用图像处理控制的测量方法，在测量过程中最大限度地降低了测量合并场的不确定度，保证了测量的质量标准。在此步骤中，将使用不同技术表面的测量来验证和改进项目前一阶段开发的方法，以便根据测量任务进行评估。在高精度定位系统控制的位置上，通过对切割单次测量和多次测量的模拟偏差来测试我们的算法的能力。作为最后的测试，我们将把我们的方法的结果与研究所现有的显微镜的拼接系统进行比较。显示高梯度的技术表面的测量通常包括由于数据处理错误或光照不足而丢失的数据和异常值。这些测量值通常被内插以完成表面数据集。插值的测量点总是取决于其周围环境、所使用的插值方法和测量方法。通过调整被测表面的光照（强度、光谱），重新测量和随后的数据融合，可以提高测量质量。因此，通过图像处理控制测量，可以大大减少假测点的数量。由图像处理控制的测量概念也可以应用于重叠表面面积的测量，以便后续使用图像处理技术进行图像配准和融合。先前的测量是用来估计理想的平移距离，以允许图像配准与最小的配准误差。这可以通过使用总体测量（使用比预期测量更低的分辨率和更大的测量区域）和当前高分辨率测量的外推来完成。通过这些技术，我们希望在保证测量和定位不确定度的情况下实现拼接测量的自动化。

The overarching goal of the project is the automation of overlapping measurements of technical surfaces using image processing microscopes and subsequent fusion for the creation of high resolution surface data sets with larger lateral dimensions and minimized measurement uncertainty (automated stitching). Methods for the reduction of the measurement uncertainty of single measurements by adaptive lighting are to be researched as well as methods for the minimization of the uncertainties of stitching (registration errors, surface deviations through data fusion). By the means of measurement controlled by image processing the uncertainty of the merged fields of measurement is minimized during the measurement to make sure the quality criteria are met.In the step, the methods developed in the previous phase of the project are to be validated and improved using measurements of different technical surfaces to allow an evaluation in dependence of the task of the measurement. Using simulated deviations on cut up single measurements and multiple measurements with positions controlled by high precision positioning systems the capabilities of our algorithms will be tested. As a final test we will compare the results of our methods with the stitching systems of the microscopes available to the institute. The measurement of technical surfaces exhibiting high gradients often includes missing data and outliers due to data processing errors or insufficient lighting. These measurements are usually interpolated to complete the surface data set. Interpolated points of measurement always depend on their surroundings, the used interpolation method and the method of measurement. By adapting the lighting (intensity, spectrum) to the measured surface and re-measuring and subsequent data fusion the quality of the measurement can be increased. Thus the amount of false measurement points can be reduced greatly by measurement controlled by image processing.The concept of measurement controlled by image processing can also be applied to the measurement of overlapping surface areas for subsequent image registration and fusion using image processing techniques. Prior measurements are to be used to estimate the ideal translation distance to allow for an image registration with minimized registration error. This is to be accomplished by using an overview measurement (using lower resolutions and larger measurement areas than the intended measurement) and by extrapolation of the current high resolution measurement. By means of these techniques we want to automatize stitching measurements with a guaranteed measurement and positioning uncertainty.