The camouflage object detection(COD) aims to identify the goals that fully integrate into the surrounding environment, and have a wide range of application scenarios. The key challenge of COD is high similarity between the goals to be recognized and the surrounding background. In this article, we have proposed a framework based on attention mechanism, searching and focusing network (FDNET), which simulates the predation process of nature. Our FDNET model uses attention mechanisms, and on the basis of enhancing the characteristics of the characteristic semantics, the interference in the background area and foreground areas is gradually refined. The experimental results show that our sfnet is better than other methods at this stage for camouflage target detection