Given the widespread prevalence of multi-view datasets containing missing values in real-world applications, incomplete multi-view clustering (IMVC) has developed into a critical research theme. By analyzing information from diverse views, IMVC seeks to alleviate the influence of missing samples across views, thus attaining better learning outcomes. Despite significant efforts made by existing methods, several challenges remain: 1) over-reliance on consistent information across views while neglecting the in-depth excavation of complementary information; 2) the lack of effective clustering guidance during the training process. In tackling these hurdles, we devised a Dual Contrastive Incomplete Multi-View Clustering method featuring clustering-oriented guidance (DCIMVC). Firstly, we design a feature-level attention fusion mechanism to dynamically integrate cross-view complementary information. Secondly, we incorporate a feature-level contrast module to strengthen the discriminative capacity of consistent information across views, while constructing clustering-oriented constraints based on sample-centroid similarity to ensure consistency between the feature space and clustering objectives. Finally, a cluster-level contrastive learning strategy is proposed to form clear clustering boundaries by strengthening intra-cluster sample similarity and inter-cluster sample dissimilarity. Experimental comparisons with cutting-edge methods have validated the predominance of our proposed approach.