Skin lesion segmentation is not only crucial for personalized patient treatment but also plays a proactive role in the rational allocation of medical resources and advancement of medical research. However, existing skin lesion segmentation models still require further exploration. In this study, we propose a Skin Lesion Segmentation Network (SkinLSNet) based on search and recognition strategies. Specifically, it employs a Pyramid Vision Transformer (PVT) as an encoder to extract multi-scale features. Subsequently, a Receptive Field Module (RFM) is utilized to expand the receptive field of these features, followed by a Layer-by-Layer Addition Decoder (LAD) to generate a Coarse Map. Additionally, we design a Multi-scale Neighborhood Complementation Module (MNCM) to enhance feature perception across scales. This process constitutes the search stage for skin lesion targets. Finally, we design a Recognition Module (RM) to fuse multi-scale features with the Coarse Map for accurate recognition of skin lesion targets. Experimental results demonstrate that our model effectively achieves accurate segmentation of skin lesion regions and exhibits strong segmentation performance on the ISIC2017 and ISIC2018 datasets.