Forest Fire Hotspot Monitoring Based On Three-Dimensional Otsu Method

The forest fire hot spot monitoring algorithm is a very critical step in the forest fire monitoring process.It can detect forest fires more quickly and accurately,detect forest fires in time,and avoid large forest fires.Traditional forest fire hotspot monitoring methods mostly use fixed thresholds,and the brightness temperature threshold for identifying forest fire hotspots is often fixed by experience.Due to the influence of geographical and seasonal environment,such methods often have missed judgments,false judgments,and delayed judgments,etc.question.Therefore,the study of forest fire hotspot monitoring algorithms with strong adaptability and better recognition effect can monitor forest fires more accurately,thereby greatly reducing the adverse effects of forest fires on human life and social economy.Because the Himawari-8 geostationary meteorological satellite has the advantages of efficient,intuitive,accurate and continuous monitoring of forest fires,this paper selects the Himawari-8 remote sensing image data as the research data.However,there are relatively few studies on the forest fire hot spot identification algorithm based on the Himawari-8 satellite,and the advantages of the remote sensing data have not been fully exploited.In this regard,based on the principle of forest fire hot spot monitoring,this paper selects the light temperature sensitive bands when the forest fire occurs,respectively:The 7-band near mid-infrared 3.9μm and the 14-band near long-wave infrared 11.2μm are used to segment the remote sensing images using the three-dimensional Otsu method to obtain potential forest fire hot spots.Shortcomings,the context method was introduced to identify potential forest fire hotspots.The algorithm can not only make full use of the time and space advantages of Himawari-8 satellite data,but also make up for the shortcomings of the existing forest fire hotspot monitoring algorithms,and promote the forest fire monitoring process that acts on other satellite remote sensing image data.The results of the study are as follows:(1)Analyze and optimize the algorithm for confirming forest fire hot spots with high accuracy and suitable for Himawari-8 remote sensing satellite.The experiment adopts the sub-pixel method,the single-channel threshold method,the multi-channel threshold method,the brightness temperature combined with the vegetation index method and the context method,a total of 5 practical forest fire hot spot identification methods,and the comparison and analysis of the five forest fire hot spot identification algorithms are used.The monitoring results of forest fire hotspots from the same remote sensing image.The results show that among the five forest fire hotspot monitoring algorithms,the monitoring results of the context method are the best.(2)Forest fire hotspot monitoring based on the three-dimensional Otsu method A forest fire occurred in Yongren County,Yunnan Province at 22:15 on April 1,2020 Beijing time.According to the three-dimensional Otsu image segmentation principle,the gray value is used as the basis for one-dimensional image segmentation.Considering the large difference between the forest fire hotspot and the background feature value,the gray mean value of the forest fire hotspot neighborhood is added to expand the one-dimensional image information to two dimensional image information,and then select the neighborhood variance gray value as the third eigenvalue,synthesize the existing 2D histogram,construct a 3D histogram,obtain an adaptive segmentation threshold,and divide the potential forest fire hot spots from the segmented from the background pixels.The size of the segmentation window is 21×21 pixels,and the study area of Yunnan Province is divided into 251 sub-areas,and the 3D Otsu image segmentation operation is performed on each sub-area.The results show that the identification thresholds of potential forest fire hotspots near the forest fire are 312.1 K and ΔT*are 312.1 K and 12.0 K.Combining the context method to confirm the fire point,not only can identify the high temperature fire point,but also can monitor the fire point.Take out a low temperature smoldering fire point not far from the high temperature fire point.(3)Based on the forest fire hotspot information monitored by the three-dimensional Otsu method,the forest fire data of Yunnan Province published by the China Forest and Grassland Fire Fighting Network and the fire point product data(Level-3)of Himawari-8 were used as the accuracy verification data..The monitoring results of the algorithm in this paper are in good agreement with China’s forest and grassland fire extinguishing network.The correct rate is 0.8,the missed detection rate is 0.09,and the comprehensive evaluation value is 0.85.In order to better strengthen the accuracy verification,the fire point product data(Level-3)of Himawari-8 on April1,2020 was selected for verification.The correct rate of the monitoring results reached 0.88,the missed detection rate was 0.16,and the comprehensive evaluation value is 0.86,indicating that the algorithm in this paper can better achieve the purpose of fire point discrimination.(4)Based on the time dimension information,test the accuracy of the algorithm at different times,select the data from March 2020 to April 2020 for the experiment,on March 30th and April 3rd,when the number of fire points is small,the algorithm omission rate has dropped by about 10%compared with the previous one.From March 31st to April 2nd,when there were a large number of fire points,the algorithm omission rate dropped by about 5%compared with the previous period.At the same time,the comprehensive evaluation value of the algorithm is improved by about 5%on the basis of the previous algorithm.In this paper,the three-dimensional Otsu method is applied to the monitoring of forest fire hotspots from Himawari-8 satellite data,which solves the problem that the original forest fire monitoring algorithm has poor adaptability in different regions,is more sensitive to small-scale forest fire hotspot events,and reduces low-temperature burns.The omission of fire spots can greatly reduce the omission and false detection of forest fire hot spots.The algorithm is also applicable to other remote sensing satellite data,which can improve the accuracy of satellite remote sensing monitoring of forest fire hotspots.The research has application value and facilitates fire prevention work.It can detect forest fires in time at an early stage and prevent large and serious forest fires.