The spatial and temporal characteristics of human activities are closely related to the function of urban land use, so the social-economic function of the urban parcel can be inferred by the spatial aggregation and dispersion of human activities. Cell phone is the most popular communication terminal equipment and the distribution of cell phone users is able to reflect the distribution of population accurately. Local basic service (LBS), which is acquired from residents' cell mobile data, is constantly emerging and make it possible to achieve spatial and temporal coverage and meticulous monitoring of urban people's activities. Therefore, the mobility data of cell phone users have the potential to infer the land use function of the urban parcels. In this paper, the call detail records(CDRs), will be adopted to cluster the urban land use patterns. Firstly, the clustering characteristics of call aggregation for local scale are extracted, then a spectral clustering recognition method for urban land use is proposed. Taking Wuhan as an experimental area, the average accuracy of the method for urban land use identification is 54.6%, and the results show that this method has advantages in urban land use identification.