Fire Monitoring Method Based On Multi-temporal And Spatial Radiation Differences Of Himawari-8 Satellite

The Himawari-8 satellite is a meteorological satellite used for observing the Asia-Pacific region in the geostationary orbit.It carries an “Advanced Himawari Imager(AHI)”,which has a resolution of 0.5 km and 2 km in the visible and infrared bands,respectively,and more spectral bands in the infrared range.The observation interval is 10 minutes,which makes this sensor important for fire monitoring.However,the current fire detection method based on Himawari-8 has a large number of missed small fires and false alarms,which limits its practical application.To improve the accuracy and stability of the Himawari-8 AHI sensor for fire detection,this thesis constructs a global fire monitoring algorithm based on the multitemporal and spatial radiation differences.The proposed algorithm is compared and validated in different types of areas using the MODIS active fire product,the Himawari-8 WLF(Wildfire Detection and Location)fire product,and high spatial resolution data.The main work of the thesis is as follows:(1)Based on the information of multi-spatial radiation differences,a weighted spatial context fire monitoring algorithm is proposed.Unlike traditional context methods that only use the 3.9μm and 11μm bands to construct a serial test condition,this algorithm fully explores the role of each band in fire monitoring in the AHI sensor,and uses different bands and band combinations to construct multiple test conditions based on spatial radiation differences.The effectiveness of each test condition is calculated using a priori database,and different weights are assigned accordingly to calculate the fire probability of each pixel.The results show that the weighted spatial context method has about 12% and 63% lower commission error and omission error rates,respectively,compared to the MODIS fire products,and has significantly improved detection results compared to traditional spatial context methods and WLF fire products.(2)A multi-temporal contextual fire monitoring algorithm is proposed to reduce the number of false fire detections by combining the differences in temporal distribution of radiation with spatial radiation differences.The multi-temporal contextual method is based on the strong spatial correlation between fire pixels and surrounding pixels,and uses historical data within a certain time range to construct a time series model.Regression analysis is used to fit the time series data,predict the brightness temperature value of the target pixel,and detect fire pixels by comparing the predicted and actual values.The results show that the multitemporal contextual algorithm effectively reduces the false detection of fire pixels in the traditional contextual algorithm.Meanwhile,for the weighted spatial context algorithm,the commission error rate and omission error rate of the multi-temporal context algorithm are reduced by about 0.5% and 2%,respectively.The reduction of the missing rate is mainly due to the omission errors of small fire points by the spatial context method.