Research And Application Of Parallel Algorithm Of PM2.5 Retrieval Using Satellite Remote Sensing Data

Recently, people paid more and more attention to the problem of air-quality and air-pollution. Among those atmospheric pollutants, PM2.5 had been one of the most emphasized objects for air-pollution monitoring in China. Nowadays, issuing PM2.5 concentration was mainly depended on the real-time results ground-based sites monitored. With the development of space industry, the technology of satellite remote sensing had gradually been an effective method of monitoring air pollution. To carry out air quality assessment and climate change research often relied on real-time, extensive and long-term monitoring statistics, however the existing atmospheric inversion methods were difficult to meet the immediate requirements of the various sectors of the industry. Through the thorough analysis and research, the reasons of the problem mentioned above were summarized as the following three points: 1) In atmospheric remote sensing field, quantitative retrieval models were exploited to retrieve the satellite images to obtain the parameter — PM2.5, which was relied on some auxiliary files. To construct these auxiliary files by traditional retrieval algorithm needed long solving time and higher cost. 2) For a large number of satellite images, employing traditional inversion algorithm to retrieve these massive statistics had led to time-consuming and low efficiency problems. 3) Traditional sites-based monitoring and desktop application systems could not achieve the requirements that the public had more urgent needs to know the concentration of PM2.5 around themselves.To solve these problems, on the basis of traditional processing method, parallel processing technology was employed to give priority to solve rapid inversion of atmospheric parameters. Based on Android platform and multi-satellite data, PM2.5 pollution monitoring system is designed. The main contents this paper conducted were summarized as the following three points:1) Proposed a dynamic distribution algorithm based on task quantity and the number of work nodes.Described the fundamental theory of constructing the lookup file PM2.5 retrieval depended on in detail. In view of the characteristics of interaction between the master node and work nodes in MPI, a dynamic allocation based on task quantity and the number of work nodes was designed, which realized fastgeneration of lookup file.2) Designed a dynamic pixel feature-based segmentation task strategy. Considering pixel feature recognition module in retrieval algorithm, a dynamic segmentation algorithm based on pixel feature was proposed on the basis of parallel processing, which speeded up the efficiency from original remote sensing images to target products.3) Designed a PM2.5 monitoring system based on Android and remote sensing data. Using multi-satellites' remote sensing products, a system aiming at monitoring PM2.5 pollution based on Android platform was designed. The message pushing mechanism between Server and Client was employed to implement a high precision, large scale of alerting PM2.5 concentration in real-time.