文章基本信息

标题：Using Dempster-Shafer Evidence Theory and Choquet Integral for Image Segmentation in Gas-Liquid Two-Phase Flow
本地全文：下载
作者：Shihong Yue ; Jin Wang ; Beibei Li 等
期刊名称：International Journal of Distributed Sensor Networks
印刷版ISSN：1550-1329
电子版ISSN：1550-1477
出版年度：2009
卷号：5
DOI：10.1080/15501320802540850
出版社：Hindawi Publishing Corporation
摘要：Data fusion technique, essentially the Dempster-Shafer evidence theory, has widely been applied in image processing and significantly improving the image segmentation quality. The mass function determination in Dempster-Shafer evidence theory plays a crucial role in data fusion process. The purpose of this paper is to propose a method of automatically determining mass functions by Choquet integral and its identification algorithm. Different from the existing versions of determining the mass function values, the obtained mass function values in fact are the fuzzy measures in Choquet integral, which can be globally and accurately determined by the parameter identification algorithm of Choquet integral. Thus our proposed method is a very general and accurate one to handle the ambiguity connected with pixel classification in image segmentation, reducing the subjectivity of the existing methods. Results on real images from the two-phase flow are presented in order to verify the usefulness of our proposed method. Within this framework, we essentially investigate the possibility of determining mass functions using the concept of fuzzy clustering algorithm and the fuzzy measure in Choquet integral. Besides, in the calculation of mass functions, we have also introduced fuzzy measure information in order to fully utilize a feasible and efficient determination method that has widely succeeded in the Choquet integral application. The obtained results show the efficient and robust characters of our proposed method. The comparison research has also illustrated that the performances in image segmentation were significantly improved when using several images representing the same object. We have presented a data fusion approach to two-phase flow image segmentation based on the Dempster-Shafer evidence theory and Choquet integral tool. The electrical CT (Computed Tomography) techniques applied in the two-phase flow measurement have been becoming more and more popular, and ECT (Electrical Capacitance Tomography) and ERT (Electrical Resistivity Tomography) both are the most useful CT techniques. Comparing the real images with the reconstructed images in our experiments the fused image based on our proposed method clearly provides a better approximation than the singleton of ECT or ERT. The experimental results show that the Choquet integral tool helps improve the image segmentation quality of flow regimes greatly.