Research On Real Time Hotspot Extraction And Classification Of Himawari-8 Satellite Images

Since the industrial revolution,the process of heavy industrialization and urbanization has accelerated,and the problem of environmental pollution has become increasingly serious.Among the environmental pollution emissions,the pollutants brought by straw burning and heavy industrial production accounted for a large proportion.Since remote sensing technology can quickly and accurately detect and classify abnormal hotspots on the ground,it reduces the manpower,material resources,and financial resources consumed by traditional manual detection,and has been widely used in the monitoring of abnormal hotspots.The Himawari-8/AHI sensor is the geostationary meteorological satellite with the highest temporal resolution,spatial resolution and spectral resolution among mainstream platforms.At present,a few scholars use Himawari-8/AHI remote sensing data for hotspot monitoring and real-time classification.And directly applying the hotspot monitoring algorithm of polar orbiting satellites to the Himawari-8 platform is not effective,and does not make full use of the high temporal resolution of the Himawari-8 satellite imaging every 10 minutes.Therefore,this article will combine space-based or time-based Based on the advantages and disadvantages of the current hotspot monitoring algorithm,making full use of the advantages of the Himawari-8 satellite,a spatio-temporal context hotspot monitoring and classification algorithm and a matching cloud mask method are proposed.The main work of this paper is as follows:(1)A multi-threshold cloud masking method based on Himwari-8 is designed.The advantages and disadvantages of current cloud masking methods are summarized,and a multi-threshold cloud masking algorithm suitable for Himwari-8 remote sensing data is proposed.The experimental area is selected in Southwest China and North China with water bodies and land.The extracted cloud mask is compared with MODIS cloud products,and the difference between the two cloud masks is compared qualitatively and quantitatively by means of visual interpretation and accuracy evaluation,and the reasons for the difference in accuracy are analyzed.The multi-threshold cloud mask algorithm proposed in this paper has a good performance in identifying thick clouds and high clouds,but is relatively poor in identifying thin clouds.The overall classification accuracy and false detection rate of the algorithm are 86.7% and 19.8%,respectively.(2)A hotspot monitoring algorithm suitable for spatio-temporal context based on Himwari-8 is proposed.Through a large number of literatures,I have read and understood the shortcomings of the time-based or spatial information-based hotspot monitoring algorithms of the current mainstream satellite platforms: the background temperature solution based on spatial information,due to the geographical properties of the target pixel and the surrounding background pixels on the ground difference,the radiation brightness temperature of the background pixels around the target cannot directly replace the actual brightness temperature of the target pixel;The lack of historical background data eventually leads to inaccurate solution of background temperature and low hotspot recognition rate.However,the Himawari-8 satellite can provide continuous satellite data with higher spatial resolution,which provides an important data source for hotspot monitoring combined with the two algorithms.Based on the Himwari-8 geostationary satellite data,this paper combines the spatial and temporal information of pixels to propose a spatiotemporal context hotspot monitoring algorithm.The algorithm was used to monitor the Chongqing area of China,and the monitoring results were verified with MODIS standard hot spot products and compared with traditional time-based or spacebased hotspot identification algorithms to analyze the causes of errors,and to identify the missing FRP fire analysis.Compared with the traditional time-based and space-based fire detection algorithm methods,the spatio-temporal context algorithm has higher accuracy in hot spot detection.Compared with the standard context algorithm,the missed mention rate and false mention rate are reduced.This demonstrates that continuous-time information about observed targets can improve the accuracy of hotspot identification.(3)A real-time classification and monitoring algorithm for hotspots based on Himwari-8 is proposed.By establishing a sample database,and then using the space,time,and brightness temperature characteristics of the hotspot itself,the monitored hotspots are classified in real time.Using the algorithm in this paper,the hotspot implementation classification is carried out in North China,and then two methods are designed for accuracy verification.Based on the object-oriented analysis of nearly 700 industrial heat sources,this paper finds that different types of industrial heat sources have unique temperature characteristics,and according to their thermal characteristics,industrial heat sources are summarized into four categories: cement plants,ferrous metals,and coal chemical plants category,oil and gas processing category.Classification of hotspots in North China shows that ferrous metals are mainly concentrated near Tianjin and Tangshan,most of the coal chemical plants are located in the south of Shanxi Province,and the hotspots of biomass combustion are more distributed in parts of Inner Mongolia and Shanxi.The classification accuracy of this algorithm is 84.61%.Experimental and application results show that the hotspot monitoring and trial classification algorithm based on Himawari-8 spatio-temporal context proposed in this paper has high efficiency in cloud masking,hotspot identification and hotspot classification.It has certain practical value for the environmental monitoring and the adjustment and optimization of industrial structure in our country.