Traditional fire detection methods use the high temperature emission characteristics in mid or thermal infrared bands of the MODIS or AVHRR data to extract burning area. It is very hard for these methods to identify small fire regions such as sub-pixel due to the limitation of spatial resolution. Recently researchers have found that shortwave infrared (SWIR) data can also be used to identify and detect high temperature targets. Compared with the thermal infrared data, SWIR has a big discrimination against different features with different temperature. Thus it can identify accurately the location of high temperature targets. In this paper, we acquired fire point products by using Landsat-8 OLI data which has spatial resolution up to 30 m. The main procedure includes two steps. The improved Normalized Burning Ratio Short-wave(NBRS) is calculated at first to adaptively acquire suspected fire points based on the spectral characteristics of fire points in the near infrared and shortwave infrared. Then most false positive points are excluded based on the relationship between peak value in shortwave infrared band of fire points. This algorithm is capable of detecting the burning area around 10% in one pixel. With the premise of avoiding the interference of cloud and constructions, it can also keep a nearly 90% accuracy and low missing rate around 10%.