Vegetation coverage represents an important role in lessening soil loss, protecting environment and improving the standard of living.USLE, as a convenient model, is applied widely.The studies mainly on method extracting vegetation coverage by using spatial and temporal remote sensing data.As a result of mixing pixels, there is something itself shortcoming for the technique of NDVI monitoring vegetation coverage, and it needs much manpoer and time for field sampling.Linear spectral unmixing is a simpler way reducing workload and improving accuracyof extracting vegetation coverage in remo te sensing image.It need not field sampling and regressive analysis.The value extent of vegetation coverage factor C in USLE model lies between 0 and 1, but the value changed greatly.So the precision of C value is a key to USLE model.