The discovery and management of network assets are crucial for network security. Faced with the vast IPv6 address space, traditional IP scanning methods are no longer applicable. Existing IPv6 address prediction algorithms suffer from low hit rates and insufficient sub-prefix coverage. Therefore, the paper proposes 6DoubleTree, an IPv6 address prediction algorithm that adopts a prefix and suffix partition tree structure. 6DoubleTree divides IPv6 addresses into prefix (high 64-bit) and suffix (low 64-bit) parts, and performs IPv6 address pattern mining and refinement on the BGP prefix space to generate IPv6 target addresses. Comparative experiments were conducted in a real IPv6 network environment, and experimental results on 3 IPv6 seed sets of 25k, 50k, and 0.1M showed that under the same budget conditions, compared to 7 algorithms based on seed address structural information, such as 6Forest and 6Graph, 6DoubleTree averagely increased the number of active addresses by 54.92% and the number of active /64 prefixes by 60.25%.