中文摘要：

针对现有的基于线性网络编码的网络拓扑推断算法中拓扑错误推断率较高、节点资源消耗大等问题，提出了一种改进的网络拓扑推断解决方案。在无链路丢包的情况下，提出了组合包测量方法，不仅适合二叉树类型的网络，也适合于存在节点度数大于3的中间节点的树状网络拓扑；在存在链路丢包的情况下，提出了快速发包算法，避免了网络中间节点和探测包帧结构设计上的额外开销。仿真结果表明，该算法具有更广泛的适用范围，具有更严谨的推断过程，同时能够在不增加网络节点运行负担的情况下降低错误推断率。

英文摘要：

The existing topology inference using linear network coding had some disadvantage such as high wrong inference and nodes resources consumption. In this paper, we presented an improved algorithm of topology inference based on network tomography with network coding. In lossless tree, a combination-packets measurement method, which was fit not only for binary trees but also for general trees with three or more degrees of nodes, was presented. The fast-sending packets method was presented in lossy trees, to avoid the extra resources consumption in intermediate node and the frame design of probe packets. Simulation results showed that the improved algorithm can be applied to many kinds of multi-tree network to inference the network topology more rigorously and efficiently without increasing the extra resources consumption of nodes.