Method for automatically identifying plumes of large point sources of emissions around the world based on satellite data

Abstract

The application provides a method for automatically identifying smoke plumes of global large point sources based on satellite data, which is based on the column concentration data retrieved by high-resolution satellite remote sensing, selects the region around the global large point source as the research region, and obtains the local concentration enhancement data by deducting the background concentration from the original column concentration. The best global segmentation threshold in the research region is obtained, and the median of the foreground pixels is further taken as the final threshold. The pixels greater than the final threshold in the concentration enhancement image are recorded as the final pixel group, and the information of the final pixel group is traversed to screen out the initial screening plume pixels that meet the conditions, so as to screen out the plume pixels that meet the conditions. The retrieved global large point source emission plume is artificially identified, and the false positive plume is removed, so that the global large point source pollution gas emission smoke plume is finally obtained. The application can automatically identify the emission plume of the preset global emission point source during satellite observation.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of atmospheric environmental pollution monitoring, and particularly relates to a method for automatically identifying smoke plumes of global large point sources based on satellite data. BACKGROUND

[0002] Improving and maintaining good atmospheric environmental air quality is one of the key tasks of ecological environment protection in China. Precise and real-time pollution emission monitoring of large emission sources is an important means to protect atmospheric environmental quality. Traditional ground-based remote sensing monitoring and ground chemical sampling analysis means cannot continuously and effectively observe a large area of the world, while satellite remote sensing has the advantage of continuous observation in a large range. However, satellite remote sensing observation is limited by cloud interference and other problems, and only in the case where the observation condition allows can it have the ability to identify large atmospheric pollution point sources in a long observation time. The instability of satellite remote sensing data coverage poses a great challenge to manual identification of the emission plumes of a large number of global large pollution point sources, and therefore it is urgent to develop a method for judging the emission plumes of global large point source pollution gases based on satellite observation data. SUMMARY

[0003] In order to better monitor global air pollution events and solve the problem that it is difficult for artificial to identify the emission plumes of pollution gases in satellite remote sensing images from global satellite monitoring big data, thereby increasing the monitoring intensity of global large point source pollution gases, the present application provides a method for automatically identifying the emission plumes of global large point sources based on satellite data, which is suitable for Level-2 product data inverted by officials and can automatically identify the emission plumes of global emission point sources during satellite observation.

[0004] To achieve the above-mentioned purpose, the present application adopts the following technical solutions:

[0005] A method for automatically identifying the emission plumes of global large point sources based on satellite data, comprising the following steps:

[0006] Step 1: Based on the data of high-resolution satellite remote sensing inversion of column concentration, the region around the global large point source as the center is selected as the research region, and the original column concentration is deducted from the background concentration to obtain the local concentration enhancement data eliminating the background concentration by calculating the background concentration value of each pixel according to the median filter principle;

[0007] Step 2: Based on the Otsu algorithm-maximum inter-class variance method, the best global segmentation threshold value in the research region is obtained, and the median of the foreground pixels is further taken as the final threshold value;

[0008] Step 3, record the pixels greater than the final threshold value in the local concentration enhancement image as the final pixel group, traverse the information of the final pixel group, and screen out the initial screening plume pixels meeting the conditions from the final pixel group according to the factory position, and then screen out the plume pixels meeting the conditions from the final pixels according to the position of the initial screening plume pixels; when the number of the plume pixels meeting the conditions is greater than 4, record the plume pixel group as a plume candidate;

[0009] Step 4, manually identify the retrieved global large point source plume, and eliminate false positive plumes, to finally obtain the global large point source pollution gas emission plume.

[0010] Further, in step 1, the background concentration value of each pixel is calculated by means of the median filter principle in image processing, the mean value, median value and standard deviation of the center pixel and the pixels within a certain range are calculated, and the background value of the center pixel is determined according to the relationship between the mean value, median value and standard deviation.

[0011] Further, in step 1, the local concentration enhancement image is obtained by subtracting the obtained background concentration image from the original satellite observation image.

[0012] Further, in step 2, the Otsu algorithm-maximum intra-class variance method is used to determine the global threshold value in the concentration enhancement image after background subtraction, and the pixel concentration values are divided into foreground group and background group according to the global threshold value.

[0013] Further, in step 2, the final threshold value refers to the median value of the foreground group, and the pixels greater than the median value of the foreground group are recorded as the final pixel group, which contains possible plume pixels.

[0014] Further, in step 3, the initial screening plume pixel is a pixel meeting the following conditions: adjacent to the factory pixel and the difference between the wind speed direction and the plume direction is less than a certain threshold value.

[0015] Further, in step 3, based on the information of the initial screening plume pixel, a pixel meeting the following conditions is screened out from the final pixel group: adjacent to the initial screening plume pixel and the difference between the wind speed direction and the plume direction is less than a certain threshold value.

[0016] Further, in step 3, by means of iteration, all the pixels meeting the conditions are screened out from the final pixel group as possible plume pixels, and only when the number of the plume pixels is greater than or equal to 4, the screened pixel group is recorded as a plume candidate.

[0017] Beneficial effects:

[0018] The present application is based on hyperspectral satellite observation data to automatically identify global large point source plume through data processing. Compared with other technical means (such as ground observation, walk observation, unmanned aerial vehicle observation, airborne observation), the present application has the following advantages and benefits:

[0019] High spatial coverage: satellite observation can realize global monitoring and identification. As long as there is no cloud condition, large emission point sources in urban areas or remote areas can be accurately captured. In contrast, ground observation and walk observation are limited by the location and coverage of observation sites, while unmanned aerial vehicle and airborne observation face the limitations of flight time and resource cost.

[0020] Daily monitoring capability: satellite observation data can provide near real-time data acquisition and processing, making the process of monitoring and identifying large emission point sources more efficient and timely. In contrast, ground observation and walk observation require a long time for data collection and processing, while unmanned aerial vehicle and airborne observation require researchers to operate to monitor and acquire data, which has a large time cost and labor cost.

[0021] Cross barriers: satellite observation is not limited by geographical environment and terrain, and can easily monitor large emission point sources located in mountainous areas, water areas or other difficult-to-reach areas. In contrast, ground observation and walk observation are restricted by terrain and geographical conditions, while unmanned aerial vehicle and airborne observation may be subject to environmental restrictions and flight safety considerations.

[0022] Comprehensive analysis capability: using satellite observation data, the plume of known point sources and unknown emission sources around the world can be identified comprehensively and quickly. Further, comprehensive analysis can be performed with other data sources (such as ground measurement data, meteorological data) to provide more comprehensive environmental assessment and research on emission source characteristics. Through comprehensive analysis, the pollution emission degree of large emission point sources and the impact on the atmosphere can be more accurately determined. BRIEF DESCRIPTION OF DRAWINGS

[0023] The present application will be further described below in conjunction with the drawings and examples.

[0024] Figure 1a , Figure 1b , Figure 1c is the observation result of the Tropomi satellite payload of Sentinel-5 around a certain large point source; wherein, Figure 1a is the original concentration image, Figure 1b is the background concentration image calculated, Figure 1c is the concentration enhancement image of the region;

[0025] Figure 2 is a schematic diagram of the example marked image greater than the optimal global segmentation threshold and the final threshold according to the present application;

[0026] Figure 3 is a schematic view of the possible plume of emissions generated by a large point source, obtained by applying the present application. DETAILED DESCRIPTION

[0027] The present application is a method for automatically identifying plumes of pollution gas emissions from large point sources based on satellite observation data, which is based on the spatial distribution of atmospheric pollution gas concentration data (L2 level product) observed by satellites, and determines the plume of emissions from large point sources by means of the principles of image processing methods, and specifically comprises the following steps:

[0028] Step 1: Based on the data of the column concentration retrieved by high-resolution satellite remote sensing, the area around the global large point source is selected as the research area, and the background concentration value of each pixel is calculated according to the principle of the median filter, and the original column concentration is deducted from the background concentration to obtain the local concentration enhancement data after eliminating the background concentration.

[0029] Step 2: Based on the Otsu algorithm - maximum inter-class variance method, the best global segmentation threshold value in the research area is obtained, and the median of the foreground pixels is further taken as the final threshold value.

[0030] Step 3: The pixels in the local concentration enhancement image that are greater than the final threshold value are recorded as the final pixel group, and the information of the final pixel group is traversed, the initial screening plume pixels that meet the conditions are screened from the final pixel group according to the position of the factory, and the plume pixels that meet the conditions are screened from the final pixel according to the position of the initial screening plume pixels. When the number of plume pixels that meet the conditions is greater than 4, the plume pixel group is recorded as a plume candidate.

[0031] Step 4: The retrieved plumes of emissions from global large point sources are manually identified, and false positive plumes are removed, and finally the plumes of pollution gas emissions from global large point sources are obtained.

[0032] Preferably, in step 1, the calculation of the background concentration value of each pixel is by means of the filter principle in image processing, the mean, median and standard deviation of the center pixel and the pixels within a certain range are calculated, and the background value of the center pixel is determined according to the relationship between the mean, median and standard deviation.

[0033] Preferably, in step 1, the local concentration enhancement image refers to the local concentration enhancement image obtained by subtracting the obtained background concentration image from the original satellite observation image.

[0034] Preferably, in step 2, the Otsu algorithm - maximum inter-class variance method is used to determine the best global segmentation threshold value in the concentration enhancement image after background subtraction, and the pixel concentration value is divided into foreground and background groups according to the best global segmentation threshold value.

[0035] Preferably, in step 2, the final threshold value refers to the median value of the foreground group, and the pixels greater than the median value of the foreground group are recorded as the final pixel group, and the final pixel group contains possible plume pixels.

[0036] Preferably, in step 3, the preliminary screening plume pixel is screened from the final pixel group and meets the following conditions: adjacent to the factory pixel and the difference between the wind speed direction and the plume direction is less than a certain threshold.

[0037] Preferably, in step 3, based on the information of the preliminary screening plume pixel, the pixel meeting the following conditions is screened from the final pixel group: adjacent to the preliminary screening plume pixel and the difference between the wind speed direction and the plume direction is less than a certain threshold.

[0038] Preferably, in step 3, through iteration, until all pixels meeting the conditions are screened from the final pixel group as possible plume pixels, and only when the number of plume pixels is greater than or equal to 4, the screened pixel group is recorded as plume candidates.

[0039] The application will be described in detail below with reference to the accompanying drawings and specific implementation cases. However, the following examples are only used to explain the application, and the protection scope of the application should include the entire content of the claims, and through the description of the following examples, a person skilled in the art can fully realize the entire content of the claims of the application.

[0040] Based on the spatial and temporal distribution of pollution gases observed by high-resolution satellite remote sensing, the column concentration values of pollution gases at each pixel position are used to identify large point source pollution plumes, including the following steps:

[0041] Step 1, the position of the large point source can be pre-set, and a certain latitude and longitude range around the point source position is selected as the research area, and the column concentration in the area is extracted as the original column concentration data, as shown in Figure 1a .

[0042] Step 2, calculate the background value of each pixel point in the research area: for each pixel point, calculate the mean value value mean , the median value value median and the standard deviation value std， of the concentration values of all pixels within a certain range around the pixel point according to formula (1), formula (2) and formula (3), then the background concentration of each pixel point is obtained by formula (4), and finally the background concentration value data value background is obtained, and the result is shown in Figure 1b .

[0043]

[0044] value median= median({VCD i |i∈n}) (2)

[0045]

[0046] where n represents the number of pixels within a certain range around the target pixel, VCD i represents the CO column density observed by the i-th selected satellite; median({VCD i |i∈n}) represents the median of n VCD values.

[0047]

[0048] where L is an empirical parameter, generally taken as 2.5.

[0049] Step 3, calculate the concentration enhancement data: subtract the background concentration value data from the original concentration at the corresponding position, Figure 1c as shown.

[0050] Step 4, based on the Otsu algorithm-maximum between-class variance method, calculate the optimal global segmentation threshold in the concentration enhancement data set, and divide the data in the concentration enhancement data into foreground group data and background group data. The specific steps for calculating the optimal global segmentation threshold are as follows: (1) calculate the histogram of the concentration enhancement data set, and count the frequency of each concentration value; (2) traverse each possible threshold, and at each threshold, divide the concentration values in the concentration enhancement data into foreground group and background group; (3) calculate the weight, average concentration value and intra-class variance of the foreground group and the background group respectively, and calculate the inter-class variance according to the intra-class variance and the class weight; (4) select the threshold that maximizes the inter-class variance as the optimal global segmentation threshold. For the foreground group data, select the median of the foreground group data as the final threshold in the concentration enhancement data, and mark the pixel points greater than the final threshold as the final pixel group, as shown in Figure 2 . Figure 2 In the enhanced concentration image shown, the pixel points with an enhanced concentration value greater than the global threshold are marked as black, i.e. the aforementioned foreground group pixels. Further, the pixel points greater than the median in the foreground group pixels are further marked as white, and the potential plume pixels are included in the white pixel group.

[0051] Step 5, from the pixels in the final pixel group, select the pixels that meet the following conditions: (1) adjacent to the factory pixel; (2) the difference between the plume stem direction formed by the factory pixel and the pixels in the final pixel group and the wind speed direction of the factory pixel is less than a certain threshold, and the empirical value is 30°.

[0052] Step 6, based on the information of the preliminary screened plume pixels, screen the pixels from the final pixel group that meet the following conditions: (1) adjacent to the preliminary screened plume pixel group; (2) the plume direction formed by the pixels in the final pixel group and the factory located pixel is less than the threshold value, the empirical value is 30°, from the wind direction of the factory located pixel.

[0053] Step 7, iteratively find the pixels that meet the conditions until all the pixels that meet the conditions are screened from the final pixel group as possible plume pixels, and only when the number of plume pixels is greater than or equal to 4, the screened pixel group is recorded as a potential plume, as shown in Figure 3

[0054] Step 8, for the screened potential plume, eliminate false positive plumes by manual identification, and finally obtain the target plume for discharge estimation.

[0055] Those skilled in the art will readily understand that the above description is only a preferred embodiment of the present application, and is not intended to limit the present application, and any modifications, equivalent replacements and improvements made within the spirit and principles of the present application shall be included in the protection scope of the present application.​

Claims

1. A method for automatically identifying plumes from large point sources around the world based on satellite data, characterized in that, It comprises the following steps: Step 1, based on the data of column concentration retrieved by high-resolution satellite remote sensing, the surrounding area of the global large point source as the center is selected as the research area, and the background concentration value of each pixel is calculated according to the principle of median filter, and the original column concentration is deducted from the background concentration to obtain the local concentration enhancement data eliminating the background concentration; Step 2, based on the Otsu algorithm-maximum inter-class variance method, the best global segmentation threshold in the research area is obtained, and the median of the foreground pixels is further taken as the final threshold; Step 3, the pixels greater than the final threshold in the local concentration enhancement image are recorded as the final pixel group, the information of the final pixel group is traversed, the initial screening plume pixels meeting the conditions are screened from the final pixel group according to the position of the factory, and the plume pixels meeting the conditions are screened from the final pixel according to the position of the initial screening plume pixels; when the number of plume pixels meeting the conditions is greater than 4, the plume pixel group is recorded as the plume candidate; Step 4, the plume of the global large point source emission searched is artificially identified, and false positive plumes are removed, and finally the plume of the global large point source pollution gas emission is obtained.

2. The method for automatically identifying plumes of large point sources based on satellite data according to claim 1, characterized in that, In step 1, the calculation of the background concentration value of each pixel is based on the principle of median filter in image processing, the mean, median and standard deviation of the center pixel and the pixels within a certain range are calculated, and the background value of the center pixel is determined according to the relationship among the mean, median and standard deviation.

3. The method for automatic identification of plumes from large point sources based on satellite data according to claim 1, characterized in that, In step 1, the local concentration enhancement image is obtained by subtracting the background concentration image from the original satellite observation image.

4. The method for automatic identification of plumes from large point sources based on satellite data according to claim 1, characterized in that, In step 2, the Otsu algorithm-maximum inter-class variance method is used to determine the global optimal threshold in the concentration enhancement image after background subtraction, and the pixel concentration value is divided into foreground group and background group according to the global optimal threshold.

5. The method for automatic identification of plumes from large point sources based on satellite data according to claim 4, characterized in that, In step 2, the final threshold refers to the median of the foreground group, and the pixels greater than the median of the foreground group are recorded as the final pixel group, so the final pixel group contains possible plume pixels.

6. The method for automatic identification of plumes from large point sources based on satellite data according to claim 1, characterized in that, In step 3, the initial screening plume pixel is a pixel meeting the following conditions screened from the final pixel group: adjacent to the factory pixel, and the difference between the wind speed direction and the plume direction is less than a certain threshold.

7. The method for automatic identification of plumes from large point sources based on satellite data according to claim 1, characterized in that, In step 3, based on the information of the initial screening plume pixel, the pixel meeting the following conditions is screened from the final pixel group: adjacent to the initial screening plume pixel, and the difference between the wind speed direction and the plume direction is less than a certain threshold.

8. The method for automatic identification of plumes from large point sources based on satellite data according to claim 1, characterized in that, In step 3, through iteration, all pixels meeting the conditions are screened from the final pixel group as possible plume pixels, and only when the number of plume pixels is greater than or equal to 4, the screened pixel group is recorded as the plume candidate.

Images