In heterogeneous graph recommendation tasks, the interaction data between users and items contains heterogeneous characteristics and high-order semantic information, providing abundant data support for the optimization of recommendation systems. Most existing mainstream methods construct multi-graph representations based on metapath mechanisms. However, the subjectivity and limitations in meta-path selection often lead to the introduction of noisy data or the omission of critical information, which in turn constrains the improvement of recommendation performance. To address this issue, this paper proposes a structural and semantic commonality perception heterogeneous graph contrastive learning model (SSCP-HGC) for recommendation tasks, aiming to reduce the dependence on the quality of meta-paths by leveraging both structural and semantic commonalities in heterogeneous graphs. Specifically, SSCP-HGC introduces a structural commonality enhancement module, which reinforces the structural commonality across graphs by capturing stable and implicit connections in multiple graphs. Meanwhile, a semantic commonality generation module is designed to extract core semantics from the high-order semantic information encoded by meta-paths, revealing the semantic consistency of nodes of the same type across multiple graphs. These modules jointly optimize the graph representations within a contrastive learning framework. Experimental results on multiple public datasets show that the proposed method outperforms existing methods, validating its effectiveness in reducing meta-path quality dependency and enhancing recommendation performance.