中文摘要：

气泡及气泡群动态演化行为的无干扰测量是气液两相流的研究重点和难点，双视角成像技术有望解决这一难题，然而现有双视角重建算法无法得到令人满意的结果。本研究对基于压缩感知的SD—POCS算法进行改进，使用直接反投影计算结果作为初始图像，动态调整迭代步长，使该算法适用于仅仅有两个角度投影数据的极端情况，从而实现双视角下两相流场气泡结构的三维图像重建。在此基础上，本研究建立了气泡识别算法，对重建结果进行气泡分析，虚拟实验表明，本方法能对气液体系中的气泡进行有效识别。

英文摘要：

Dynamic behavior measurement of bubble and bubbles is the emphasis and difficulty of gas-liquid two-phase flow study, and two view imaging technique is expected to solve this problem, but the existing reconstruction algorithm is not satisfying. In this study the SD-POCS algorithm based on compressed sensing was improved, and the direct back-projection result was adopted as the initial image for the further reconstruction and the iteration step was dynamically adjusted for the situation of two projections. The 3D reconstruction of bubble in two-phase flow was realized with this algorithm, and the bubble recognition algorithm was put forward for bubble analysis. The result of virtual experiment shows that the improved algorithm for two view imaging is useful for bubble reconstruction and recognition.