Blockchain technology has shown great application potential in many fields such as finance, supply chain, and the Internet of Things. As blockchain technology keeps evolving, the issue of insufficient scalability has progressively turned into a crucial bottleneck impeding its further progress. The application of sharding technology to enhance the scalability of blockchain systems has attracted considerable interest from the research community. However, the majority of sharding solutions suffer from a number of problems, including high ratio of cross-shard transactions, prolonged transaction confirmation latency, and unbalanced load among shards. To solve these problems, we propose a blockchain sharding method based on dual-constraint label propagating. This sharding method mainly depends on two constraints in the label propagation process to determine whether to perform partition transfers on blockchain accounts. Constraint 1 is to consider the correlation among accounts when splitting accounts. This constraint serves to minimize the cross-shard transactions ratio. Constraint 2 is that the load factor of the blockchain sharding system after the tag update cannot exceed the load factor before the update. This can bring the load of each shard to an almost consistent state, thereby reducing transaction confirmation latency. To verify the effectiveness of our proposed method, we compare it with the Metis and Monoxide algorithms in three aspects: cross-shard transaction ratio, transaction throughput, and transaction confirmation latency. Furthermore, we examine the influence of the load distribution across individual shards within the blockchain network on both throughput and transaction confirmation latency. The experimental results show that our proposed method makes the load among shards more balanced, thereby reducing the ratio of cross-shard transactions, decreasing the transaction confirmation latency, and increasing the transaction throughput.