Method for automatically extracting small and medium-sized burned area based on enhanced combustion index

Abstract

The invention discloses a small and medium-sized burned area automatic extraction method based on an enhanced combustion index, and the method comprises the steps: determining a fire site according to the azimuth angle of the fire site relative to the position of an observer and the visual distance between the fire site and the observer, and obtaining the images of the fire site before and after a disaster; data cloud mask processing is carried out on the obtained fire image through cloud mask identification, so that the influence of a large amount of dense clouds is removed, and then an enhanced normalized combustion index is constructed, so that a burned area presents a highlight value, and the highlight value is obviously different from surrounding background pixels; and performing an Ostu automatic segmentation algorithm on the fuzzified ENBR image data to complete the extraction of the burned area. According to the method, the data features of the burned land are automatically extracted based on ENBR and OSTU algorithms, and compared with a traditional dNBR and BAI manual threshold judgment method, the method has better extraction precision and stability, the extraction result can be analyzed, better area and threshold self-adaptive capacity is achieved, and judgment precision is high.

Description

technical field [0001] The invention relates to the technical field of fire data analysis, in particular to an automatic extraction method for small and medium-sized fire traces based on an enhanced combustion index. Background technique [0002] Forest fires seriously affect the land surface ecological environment and carbon balance, resulting in air pollution and carbon emissions, reduction of vegetation carbon sequestration, and global climate change. Estimation and other important data sets for post-disaster assessment. High-temporal-spatial resolution satellite remote sensing technology has the advantages of high efficiency and labor-saving. It is widely used in forest fire monitoring and feature extraction of burned areas. The extraction of burned areas involves spectral index, supervised classification, and time-series image segmentation. However, most of the current methods are based on supervision and semi-supervision, and it is difficult to realize automatic extrac...

AI Technical Summary

Problems solved by technology

Machine learning classification and unsupervised classification methods combined with Sentinel-1 SAR images can automatically locate the burnt area and the degree of burning. However, there are still problems such as omission of identification of small fire sites and misjudgment of plaques caused by disturbances around the fire site.

Images