In real-world agricultural scenarios, 3D plant point cloud data are often incomplete due to occlusions, environmental disturbances, and sensor limitations, impeding downstream tasks such as phenotypic analysis, growth monitoring, and digital twin construction. Most existing completion approaches neglect the potential of leveraging complete point clouds from other domains, thereby underutilizing cross-domain information. To address this limitation, we propose a cross-domain point cloud completion framework integrating a Feature Low-Rank Mapping (FLRM) module and a Dual-Frequency Domain Prompt (DFDP) module. Specifically, FLRM alleviates domain shifts by aligning source and target features within a shared low-rank space, while DFDP leverages the Graph Fourier Transform to enhance the extraction of global and local geometric structures in the frequency domain. Extensive experiments on the proposed PPCC-CD dataset demonstrate that our method outperforms existing baselines in both accuracy and robustness, achieving a 30.15% improvement in CDL₁ over the second-best approach. These results highlight the effectiveness of our approach in reconstructing high-fidelity plant structures under challenging cross-domain and data-scarce conditions.