【作者】 张凯；

【导师】 岳士弘；

【作者基本信息】 天津大学 ， 模式识别与智能系统， 2010， 硕士

【摘要】 近年来发展起来的电学层析成像技术(Electrical Tomography，简称ET)在两相流/多相流流动参数分布状况的测量方面取得了一定的进展，已经成为解决两相流参数检测的有效手段，具有很大的潜力和广阔的工业应用前景。电学层析成像技术ET (Electrical Tomography,简称ET)是目前最为广泛研究的一种过程层析成像技术，它具有系统成本低、成像速度快、适用范围广、非侵入式等优点，在PT中很有优势。因而，电学层析成像技术成为研究的热点。然而，尽管ET技术取得了较大的进展，但是图像重建的精度和速度却一直是没有获得很大的进展。为了获得更佳的图像质量和更快的成像速度，许多研究者进行了不懈的努力，提出了多种图像重建算法，如反投影法、Landweber的迭代法等，但是这些算法各有不同的适用场合与局限性。针对目前断层成像重建图像算法存在的成像精度较低和速度较慢的问题，基于模糊聚类算法，我们提出了一种具有原创性的新型图像重建算法。与现有的成像算法不同的是，我们把不同的传感器所获取的信息有序地映射到一个向量的空间进行聚类分析。同时，通过特征分组加权，充分考虑特征之间度量值的不均衡性，更好地描述了数据中不同特征作用的差异。实验证明，我们提出的算法能够充分地利用已有信息，获得更好的图像重建效果。进一步地，我们提出的算法实现简单，计算复杂度低，能够实时、动态地更新计算结果，为图像重建技术提供一个全新的手段。最后，又从数据融合的角度做了ERT和ECT图像融合的初步研究。我们应用Comsol公司开发的Comsol软件在电磁方面强大的数据仿真功能，对我们所需要的实验数据用其仿真得到。一方面可以完全得到理想中的数据，另一方面，对于我们的新算法来说，完全应用理想化下的数据有利于对算法效果的验证以及算法的发展。更多还原

【Abstract】 With the rapid development in recent several ten years, Electrical Tomography(Electrical Tomography, ET) techniques have been successfully used in two-phaseflow/multiphase flow parameter measurement, and ET techniques have great potentialand could be widely used in industry.Electrical Tomography (Electrical Tomography, ET) is one of the techniques ofPT, which is widely researched. It has advantages of low-cost, high-speed, robust andnon-intrusive, et. al. Thus, ET technology has become a hot research.However, despite the ET technology has made great progress, the accuracy andspeed of image reconstruction have not been achieved more progress. In order toobtain better image quality and faster imaging speed, many researchers have madegreat efforts on image reconstruction algorithm, such as LBP and Land Weber est. But,the applications of these algorithms have different situations and limitations.To overcome the problems of speed and precision of the existing algorithms, wecreatively present a method of image reconstruction algorithm based on fuzzyclustering. Different from the existing algorithms, we map the information fromdifferent sensors into a vector space for clustering analysis., Then, all attributes ofeach vector are not equally important in clustering process，and alternatively theeffects of some attributes are enlarged while the other is reduced. Consequently, ouralgorithm can fully utilize the existing information to obtain the better reconstructedimages. Consequently, our algorithm is easily realized, lower time complexity, andupdated in a real-time and automatic manner. Thus our algorithm contributes a newway to the image reconstruction techniques.Finally, we did the simple image fusion of ERT and ECT from the perspectiveof data fusion.In addition, we apply Comsol’s great simulation on electromagnetic data, to getour experimental data. It can get completely satisfactory data, on the other hand, toour new algorithm; the ideal data can be favorable to our algorithm verification andthe development of algorithms.更多还原