In recent years, Vision Transformers have achieved impressive results in camouflaged object detection (COD). However, two major challenges remain: limited perception capability and insufficient feature propagation in the decoder stage. Addressing these limitations is crucial for enhancing detection accuracy. To this end, we propose a novel camouflaged object detection network based on collaborative perception and a dual-stage decoder architecture, referred to as collaborative perception and dual-stage decoder network (CPDDNet). Specifically, a Coordinate-window collaborative attention (CWCA) module is introduced to jointly exploit global contextual information and local structural cues. In addition, we design a dual-stage decoder structure to enhance multi-level semantic feature fusion and spatial detail recovery. Extensive experiments on four widely-used COD datasets across sixteen evaluation metrics demonstrate that our method outperforms state-of-the-art approaches in both accuracy and robustness.