Face recognition has gained widespread use as a biometric technology. While many deep learning-based 3D face recognition techniques have achieved promising results using high-quality databases, recognizing faces on low-quality face data, often characterized by poses, occlusions, and temporal changes, remains a challenge, especially when captured with low-cost sensors. In this paper, we propose a novel end-to-end dual network using contrastive learning for 3D face recognition on low-quality data. In particular, we construct a pair of contrastive encoders with the MobileNet V2 backbone for contrastive representation learning. Furthermore, we introduce a joint loss function that combines the contrastive loss and the cross-entropy loss to facilitate joint contrastive learning and classification. Experiments show that our approach achieves state-of-the-art performance under different settings.