Vegetation is an important part of urban ecosystem; therefore timely access to vegetation coverage information is of great significance for monitoring urban ecological environment. Linear spectral mixture model (LSMM) was carried out for urban vegetation coverage extraction using medium-resolution Landsat TM remote sensing data. Meanwhile, the fuzzy c-means (FCM) method was chosen to extract vegetation coverage by improving the training sample selection method to obtain the end-member sample based on minimum noise transform (MNF), pixel purity index analysis (PPI), and N-dimensional visualization analysis. Finally, high-resolution SPOT5 remote sensing data extracted in two ways were used to carry out the accuracy test for vegetation coverage. The results showed that the correlation coefficients between the inspection data and LSMM-based and improved FCM-based data were 0.8252 and 0.9381, respectively. It indicated that the improved FCM-based method with higher accuracy can better eliminate the nonlinear effect of other elements.