Method for identifying local atmospheric pollution sources based on refined remote sensing data

Abstract

The application discloses a method for identifying local atmospheric pollution sources based on refined remote sensing data, which can directly reflect the real local atmospheric pollution source emission information, and can accurately locate the local emission pollution source by directly reading the remote sensing data, eliminating the interference caused by pollution transmission. In the method, a grid with a reasonable size is set, three labels are selected, and the sub-grid with transmission interference is screened out by comparing the sizes of the pollution values corresponding to the labels, so that the refined remote sensing data is obtained, and a method for further correcting the refined remote sensing data is given.

Description

TECHNICAL FIELD

[0001] The present application relates to the identification method of local atmospheric pollution sources, and particularly relates to a method for identifying local atmospheric pollution sources based on refined remote sensing data. BACKGROUND

[0002] With the deepening of the work of atmospheric pollution control, reducing pollution source emissions from the source becomes a more and more direct and effective means, and in order to implement this scheme, accurately knowing where the pollution source is, that is, accurately identifying the local atmospheric pollution source becomes a more and more important pollution control prerequisite, however, in the current technology, only by directly reading the atmospheric pollutant concentration remote sensing image can the local pollution source information be obtained.

[0003] However, the gas concentration information read in the atmospheric pollutant concentration remote sensing image includes: local emission pollutants and transmission pollutants. The transmission pollutants are pollutants emitted into the air from other areas and transported to the local area under the action of air flow. Due to the existence of transmission pollutants, the scheme in the prior art, that is, directly using remote sensing image information, cannot accurately describe the local emission and cannot meet the purpose of identifying the local atmospheric pollution source.

[0004] Based on the above problems, the present inventors have made in-depth analysis on remote sensing data in the hope of designing a method for identifying local atmospheric pollution sources that can solve the above problems. SUMMARY

[0005] In order to overcome the above problems, the present inventors have made intensive research and designed a method for identifying local atmospheric pollution sources based on refined remote sensing data. In this method, the transmission atmospheric remote sensing data is removed, and the remote sensing data that can directly reflect the real local atmospheric pollution source emission information is obtained. Directly reading the remote sensing data can accurately locate the local emission pollution source, eliminating the interference caused by pollution transmission. In this method, a grid with a reasonable size is set, and three labels are selected. By comparing the size of the label corresponding to the pollution value, the sub-grid with transmission interference is screened out, so as to obtain the refined remote sensing data, and a method for further correcting the refined remote sensing data is also given, thereby completing the present application.

[0006] Specifically, the purpose of the present application is to provide a method for identifying local atmospheric pollution sources based on refined remote sensing data, which comprises the following steps:

[0007] Step 1: calling a high spatial resolution remote sensing image containing a to-be-detected area, and drawing a grid in the remote sensing image, which divides the to-be-detected area into a plurality of sub-regions with fixed size;

[0008] Step 2: establishing a label frame in the high spatial resolution remote sensing image, preferably, the label frame includes a high value label frame, a medium value label frame and a low value label frame;

[0009] Step 3: Retrieve remote sensing images containing atmospheric pollutant concentrations in the area to be detected, add label boxes and grids to the remote sensing images, divide the remote sensing images into multiple grid units using the grid, and assign values ​​to each label box;

[0010] Step 4: Analyze each grid cell one by one to determine whether the information in the grid cell is normal. If it is normal, save the atmospheric pollutant concentration information recorded in the remote sensing image of the grid cell, and the grid cell is an information grid; if it is abnormal, clear the atmospheric pollutant concentration information recorded in the remote sensing image of the grid cell, and the grid cell is an empty grid.

[0011] Step 5: Repeat steps 3 and 4. Each time step 3 is repeated, retrieve a remote sensing image of the atmospheric pollutant concentration in the area to be detected from a different time frame. Convert each remote sensing image into a set of information grids and empty grids.

[0012] Step 6: Summarize all the obtained information grids; and summarize all the atmospheric pollutant concentration information in the information grids into a single frame of atmospheric pollutant concentration remote sensing image, thereby obtaining refined remote sensing data.

[0013] In step 2, the process of establishing the tag includes the following sub-steps:

[0014] Sub-step 2-1: Draw the outline of high-concentration emission areas of air pollutants in high spatial resolution remote sensing images. The high-concentration emission areas of air pollutants include cities, industrial parks, traffic emission areas and areas where polluting enterprises are located.

[0015] The outlines of multiple high-concentration emission areas of air pollutants are combined to obtain multiple preliminary high-value label outlines.

[0016] Sub-step 2-2: Draw the outline of the low-concentration emission area of ​​air pollutants in the high spatial resolution remote sensing image. The low-concentration emission area of ​​air pollutants includes farmland, desert and forest areas.

[0017] The outlines of multiple low-concentration emission areas of air pollutants are combined to obtain multiple preliminary median label outlines.

[0018] Sub-steps 2-3 involve drawing the outline of areas free from air pollutant emissions in a high spatial resolution remote sensing image. These areas include lakes, reservoirs, and the sea.

[0019] The outlines of multiple areas with no air pollutant emissions are combined to obtain multiple preliminary low-value label outlines.

[0020] Sub-steps 2-4: Select a rectangular region in the label outline. The side length of this rectangular region is an integer multiple of the side length of the smallest pixel in the remote sensing image of atmospheric pollutant concentration.

[0021] Through the above process of taking the union set, each sub-region has at least one preliminary high-value label profile and at least one preliminary median label profile.

[0022] In sub-steps 2-4, the rectangular region is completely within the label outline. When the label outline is too small to completely cover the smallest pixel, the label outline information is discarded and deleted.

[0023] In step 3, the process of assigning values ​​to the label box is as follows: read the pollutant concentration values ​​represented by all pixels falling within the label box in the remote sensing image, and take their average value as the pollutant concentration value of the label box, thus obtaining a complete label.

[0024] In step 4, within a grid cell, if the value of any high-value label is greater than the value of any median label and also greater than the value of any low-value label, the information in that grid cell is considered normal; otherwise, the information in that grid cell is considered abnormal.

[0025] The process of obtaining refined remote sensing data in step 6 includes the following sub-steps:

[0026] Sub-step 6-1: Group the grid cells representing the same sub-region together, extract the pixels at the same position in each grid cell, take the average value to obtain a new pixel, repeat the process of extracting pixels and taking the average value multiple times, traverse each pixel in each grid cell, and then obtain a new grid cell composed of new pixels.

[0027] Sub-step 6-2 arranges all the new grid cells according to their actual positional relationships to obtain refined remote sensing data.

[0028] The method also includes the following step 7:

[0029] Step 7: Add the grid from Step 1 and the label boxes from Step 2 to the refined remote sensing data obtained in Step 6, filter out areas with high pollution values ​​outside the label boxes as areas to be processed, and further analyze whether there are local pollution sources in the areas to be processed, and update the refined remote sensing data accordingly.

[0030] In step 7, after obtaining the region to be processed, the following sub-steps are specifically performed:

[0031] Sub-step 7-1: Retrieve the remote sensing image and its label information obtained in step 3, and add the area to be processed and the grid to the remote sensing image; retrieve the near-surface wind information of the area to be detected;

[0032] Sub-step 7-2: Compare the mean concentration corresponding to the area to be processed with the observed concentration. When the observed concentration is less than half of the mean concentration, add a mark to the area to be processed. In the refined remote sensing data, the average concentration of atmospheric pollutants corresponding to the pixels in the area to be processed is recorded as the mean concentration. In the remote sensing image retrieved in sub-step 7-1, the average concentration of atmospheric pollutants corresponding to the pixels in the area to be processed is recorded as the observed concentration.

[0033] Sub-step 7-3: When the observed concentration is not less than half of the mean concentration, analyze whether the air pollutants emitted at the high-value label in the grid cell where the area to be treated is located can be moved to the area to be treated by the wind; if so, add a label to the area to be treated.

[0034] Sub-step 7-4: Repeat sub-steps 7-1, 7-2, and 7-3 in sequence. Each time sub-step 7-1 is executed, select a different remote sensing image with different atmospheric pollutant concentrations until all remote sensing images used in step 3 have been traversed. Count the number of labels N for each area to be processed and the number of times M in which the grid cell of each area to be processed is identified as an information grid. Preferably, when the value of N is greater than 70% of the value of M, reassign the atmospheric pollutant concentration to the area to be processed.

[0035] The process of reassigning atmospheric pollutant concentration is as follows: read the markers added to the area to be processed, retrieve the remote sensing images used in sub-step 7-1 corresponding to each marker, read the atmospheric pollutant concentration information in the area to be processed in these remote sensing images, take the average value as the new atmospheric pollutant concentration of the area to be processed, and update the refined remote sensing data accordingly.

[0036] The beneficial effects of this invention include:

[0037] (1) The method for identifying local air pollution sources based on refined remote sensing data provided by the present invention abandons the traditional scheme of directly judging local pollution sources based on satellite remote sensing monitoring results of air pollution sources, eliminates the interference caused by pollution transmission in the traditional scheme, improves the accuracy of local pollution source identification and judgment, and provides reliable technical support for environmental governance.

[0038] (2) According to the method for identifying local air pollution sources based on refined remote sensing data provided by the present invention, a correction scheme is further set on the basis of obtaining refined remote sensing data, which further improves the identification accuracy of local pollution sources. Attached Figure Description

[0039] Figure 1 An overall logic diagram of a method for identifying local air pollution sources based on refined remote sensing data according to a preferred embodiment of the present invention is shown.

[0040] Figure 2 This diagram illustrates the area to be monitored after a grid has been drawn in an embodiment of the present invention.

[0041] Figure 3 This illustrates a remote sensing image of atmospheric pollutant concentration in an embodiment of the present invention;

[0042] Figure 4 A schematic diagram of a grid cell and its label in an embodiment of the present invention is shown;

[0043] Figure 5 This illustrates the refined remote sensing data obtained in an embodiment of the present invention;

[0044] Figure 6 Show Figure 3 A magnified view of the 46th grid cell;

[0045] Figure 7 Show Figure 5 A magnified view of the 46th grid cell;

[0046] Figure 8 High-resolution ground feature imagery is shown at grid cell 46. Detailed Implementation

[0047] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Through these descriptions, the features and advantages of the present invention will become clearer and more apparent.

[0048] The term “exemplary” as used herein means “serving as an example, embodiment, or illustration.” Any embodiment illustrated herein as “exemplary” is not necessarily to be construed as superior to or better than other embodiments. Although various aspects of embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless specifically indicated otherwise.

[0049] According to the present invention, a method for identifying local air pollution sources based on refined remote sensing data is provided, such as... Figure 1 As shown, the method includes the following steps:

[0050] Step 1: Retrieve a high spatial resolution remote sensing image containing the region to be detected, and draw a grid in the remote sensing image. The grid divides the region to be detected into multiple sub-regions of fixed size.

[0051] The selection of the area to be tested is generally based on administrative divisions, which can be national, provincial, or municipal administrative regions. For example, Shandong Province can be selected as the area to be tested. In other words, the area to be tested can be selected and set according to specific task requirements.

[0052] The high spatial resolution remote sensing images described in this application are satellite photographs obtained from the GF1, GF2, GF6 and other series of satellites.

[0053] The grid refers to a dense network pattern composed of intersecting horizontal and vertical lines, containing multiple adjacent cells, which are referred to as grid units in this application. The grid units are located on a remote sensing image or imagery. The sub-region refers to a specific real-world area on the ground. Each grid unit corresponds to a sub-region.

[0054] By drawing a sufficiently large grid with enough grid cells, the remote sensing image of the entire area to be detected is located within the grid, that is, the grid completely covers the remote sensing image of the area to be detected.

[0055] The size of the grid should not be too large, and the number of grid cells in the grid should not be too many. Each grid cell should contain at least a portion of the remote sensing image of the area to be detected, that is, each sub-region should contain at least a portion of the area to be detected.

[0056] Preferably, the sub-region is a square region; more preferably, the side length of the square region is 50 km. In this application, by setting this grid and obtaining the sub-regions, subsequent filtering is performed on a sub-region / grid unit basis. By setting the most suitable sub-region size, the subsequent filtering is most accurate and reasonable, ensuring the removal of abnormal data while minimizing workload. Simultaneously, a reasonable sub-region size facilitates subsequent label selection and comparison. The sub-region size provided in this application ensures that each sub-region contains a suitable number of various labels, facilitating quick label selection, assignment, comparison, and judgment.

[0057] Step 2: Create label boxes in the high spatial resolution remote sensing image. Preferably, the label boxes include high-value label boxes, medium-value label boxes, and low-value label boxes.

[0058] In step 2, the process of creating the label box includes the following sub-steps:

[0059] Sub-step 2-1: Draw the outline of high-concentration emission areas of air pollutants in high spatial resolution remote sensing images. The high-concentration emission areas of air pollutants include cities, industrial parks, traffic emission areas and areas where polluting enterprises are located.

[0060] Preferably, by reading high spatial resolution remote sensing images, cities, industrial parks, and traffic emission zones in the area to be detected are identified and their outlines are marked. This process is automatically completed by the system recognizing high spatial resolution remote sensing images. When the high spatial resolution remote sensing images are retrieved, the outlines can be directly marked. However, the outlines may not be accurate enough, which will affect the accuracy of subsequent processing and lead to errors in the final refined remote sensing data.

[0061] Based on this, this application further filters and corrects areas with high concentrations of air pollutants by performing the following operations: retrieving the land cover type information of the area to be detected, and drawing the outlines of the city and industrial park in a high spatial resolution remote sensing image accordingly; the land cover type information mentioned in this application is public information, including a map that records the type of each area and the boundaries between each area.

[0062] The enterprise's pollution discharge permit registration information is retrieved, and the outline of the location of the pollution discharge enterprise is drawn in a high spatial resolution remote sensing image based on this information; the enterprise's pollution discharge permit registration information is also public information, including the specific location and area of ​​the pollution discharge enterprise.

[0063] By combining the outlines of multiple high-concentration emission areas of air pollutants, several preliminary high-value label outlines are obtained.

[0064] In this application, when the position of the contour line cannot be accurately determined during the process of drawing the contour, the uncertain area can be discarded and the contour can be drawn based on the determinable part. Since this application adopts a scheme of simultaneously drawing contours using multiple methods and taking the union of the results, it can ensure the accuracy of the final result while improving efficiency.

[0065] Furthermore, in this application, based on multiple contour information obtained by different means, considering that contours may contain errors, and taking into account the role of contour information, this application performs union processing on overlapping or intersecting contours of the same type, which can ensure the smooth progress of subsequent label box processing and also ensure the accuracy of the final refined remote sensing data.

[0066] Sub-step 2-2: Draw the outline of the low-concentration emission area of ​​air pollutants in the high spatial resolution remote sensing image. The low-concentration emission area of ​​air pollutants includes farmland, desert and forest areas, which may be interspersed with scattered rural residences.

[0067] Preferably, by reading high spatial resolution remote sensing images, farmland, desert, and forest areas in the area to be detected are identified and their outlines are marked. This process is automatically completed by the system recognizing high spatial resolution remote sensing images. When the high spatial resolution remote sensing images are retrieved, the outlines can be directly marked. However, the outlines may not be accurate enough, which will affect the accuracy of subsequent processing and lead to errors in the final refined remote sensing data.

[0068] Based on this, this application further filters and corrects the low-concentration emission areas of air pollutants by performing the following operations: retrieving the land cover type information of the area to be detected, and drawing the outlines of farmland, desert and forest in the high spatial resolution remote sensing image accordingly;

[0069] By combining the outlines of multiple low-concentration emission areas of air pollutants, several preliminary median label outlines are obtained.

[0070] Sub-steps 2-3 involve drawing the outline of areas free from air pollutant emissions in high spatial resolution remote sensing images. These areas include water bodies such as lakes, reservoirs, and oceans.

[0071] Preferably, by reading high spatial resolution remote sensing images, lakes, reservoirs, or sea areas in the region to be detected are identified and their outlines are marked. This process is automatically completed by the system recognizing high spatial resolution remote sensing images. When the high spatial resolution remote sensing images are retrieved, the outlines can be directly marked. However, the outlines may not be accurate enough, which will affect the accuracy of subsequent processing and lead to errors in the final refined remote sensing data.

[0072] Based on this, this application further filters and corrects areas without air pollutant emissions by performing the following operations: retrieving the land cover type information of the area to be detected, and drawing the outline of the location of the lake, reservoir or sea in the high spatial resolution remote sensing image accordingly.

[0073] The outlines of multiple areas with no air pollutant emissions are combined to obtain multiple preliminary low-value label outlines.

[0074] In a preferred embodiment, during the union process, overlapping regions are first merged into a single contour region. Alternatively, two closely spaced contours without overlap can be merged into a single contour region. Here, "closely spaced" is a relative concept and needs to be determined based on actual needs. Through the above union process, each sub-region corresponds to at least one preliminary high-value label contour and at least one preliminary median label contour. Furthermore, each sub-region ultimately corresponds to at least one high-value label box and at least one median label box. Preferably, the number of preliminary high-value label contours in each sub-region is generally less than 3, and the number of preliminary median label contours in each sub-region is generally less than 5.

[0075] When other types of contour regions exist between two contour regions of the same type, these two contour regions of the same type cannot intersect. Furthermore, they cannot intersect when the minimum distance between two contour regions of the same type is greater than 7 km. For example, the contours of a high-concentration air pollutant emission area and a low-concentration air pollutant emission area are contour regions of different types and cannot intersect.

[0076] Preferably, when a portion of the high-value label box is located in one sub-region and another portion is located in another sub-region, it is also considered that there is a high-value label box in both sub-regions.

[0077] Sub-steps 2-4 involve selecting a rectangular region within the label outline. The side length of this rectangular region is an integer multiple of the side length of the smallest pixel in the remote sensing image of atmospheric pollutant concentration. The rectangular region located within the high-value label outline is the high-value label box, the rectangular region located within the medium-value label outline is the medium-value label box, and the rectangular region located within the low-value label outline is the low-value label box. In this application, setting the label box as a rectangle and limiting its side length ensures that the smallest pixel falls precisely within this rectangular region, facilitating subsequent value assignment.

[0078] Preferably, the rectangular area is completely within the label outline. When the label outline is too small to completely cover the smallest pixel, the label outline information is discarded and deleted.

[0079] The label profiles include high-value label profiles, medium-value label profiles, and low-value label profiles.

[0080] Preferably, when taking the union of the contours and setting the rectangular regions, the size of the rectangular regions should be controlled as much as possible so that the rectangular region size of each label box can satisfy 3-4 pixels falling within it. The applicant found that by setting the label boxes and each label box corresponding to 3-4 pixels, the number and size of the label boxes obtained are just right. The accuracy of subsequent judgment on whether the grid cells are normal or not is the highest, and the obtained refined remote sensing data is also more accurate.

[0081] Step 3: Retrieve remote sensing images containing atmospheric pollutant concentrations in the area to be detected, add label boxes and grids to the remote sensing images, divide the remote sensing images into multiple grid units using the grid, and assign values ​​to each label box;

[0082] The addition in this application involves drawing the same label boxes and grids at the same locations in the remote sensing image. Since the ground contours corresponding to the remote sensing image and the high spatial resolution remote sensing image are consistent, the addition process is simple and easy to operate.

[0083] The remote sensing imagery of atmospheric pollutant concentrations can be any type of remote sensing imagery characterizing atmospheric pollution, such as tropospheric NO2 column concentration, tropospheric HCHO column concentration, or near-surface O3 concentration. Preferably, in practical work, one type of pollution source information is selected for analysis and processing, and the resulting refined remote sensing data only reflects the actual local emissions of that pollution source.

[0084] When the remote sensing image of atmospheric pollutant concentration is the tropospheric NO2 column concentration, the pollutant concentration value specifically refers to the tropospheric NO2 column concentration value; similarly, when the remote sensing image of atmospheric pollutant concentration is the tropospheric HCHO column concentration, the pollutant concentration value specifically refers to the tropospheric HCHO column concentration value; when the remote sensing image of atmospheric pollutant concentration is the near-surface O3 concentration, the pollutant concentration value specifically refers to the near-surface O3 concentration value.

[0085] The remote sensing images of atmospheric pollutant concentrations, such as Figure 3 As shown in the image, each visible pixel is the smallest pixel in the remote sensing image, and different colors of the pixels represent different pollutant concentration values; that is, each pixel corresponds to a specific pollutant concentration value, which represents the pollutant concentration in the area corresponding to that pixel. When retrieving this remote sensing image of atmospheric pollutant concentration, the retrieved data includes both the location and corresponding color of each pixel, as well as the specific pollutant concentration value corresponding to each pixel.

[0086] For example, when the remote sensing image of atmospheric pollutant concentration is the concentration of NO2 column in the troposphere, the actual ground size corresponding to the smallest pixel is 3.5*7KM.

[0087] Preferably, the process of assigning values ​​to the label boxes is as follows: read the pollutant concentration values ​​represented by all pixels falling within the label boxes in the remote sensing image, and take their average value as the pollutant concentration value of the label box, thus obtaining a complete label; the label obtained after assigning values ​​to the high-value label boxes is called a high-value label, the label obtained after assigning values ​​to the median label boxes is called a median label, and the label obtained after assigning values ​​to the low-value label boxes is called a low-value label.

[0088] More preferably, when the geometric center of the pixel falls within the label frame / rectangular area, the pixel is considered to fall within the rectangular area. Since the acquisition time of each remote sensing image of atmospheric pollutant concentration is different and there are small deviations between them, the assignment of the label frame is determined by calculating the geometric center position of the pixel to determine whether the pixel falls within the rectangular area. This can eliminate the influence of the deviation between the remote sensing images of atmospheric pollutant concentration, and ensure the accuracy of the label and the accuracy of the final refined remote sensing data.

[0089] Step 4: Analyze each grid cell one by one to determine whether the information in the grid cell is normal. If it is normal, save the atmospheric pollutant concentration information recorded in the remote sensing image of the grid cell, and the grid cell is an information grid; if it is abnormal, clear the atmospheric pollutant concentration information recorded in the remote sensing image of the grid cell, and the grid cell is an empty grid.

[0090] In step 4, within a grid cell, if the value of any high-value label is greater than the value of any median label and also greater than the value of any low-value label, the information in that grid cell is considered normal; otherwise, the information in that grid cell is considered abnormal.

[0091] This application sets up grid cells and three types of labels within them, and further judges based on the size of the label values. This allows for the filtering out of grid cells with abnormal information, thereby identifying the presence of pollutant migration in those grid cells and discarding them. This process filters out grid cells without pollutant migration interference, providing a more accurate data foundation for subsequently obtaining local pollution source emission information.

[0092] Step 5: Repeat steps 3 and 4. Each time step 3 is repeated, retrieve a remote sensing image of the atmospheric pollutant concentration in the area to be detected from a different time frame. Convert each remote sensing image into a set of information grids and empty grids.

[0093] During the repetition of step 3, multiple frames of remote sensing images containing atmospheric pollutant concentrations in the area to be detected are retrieved sequentially over a given period of time. This period can be selected and set according to actual mission requirements; such as one year, one quarter, one month, or one week, etc. Since current remote sensing satellites can provide at least one frame of remote sensing image of the same area almost every day, a period of one week can provide at least 7 frames of remote sensing images.

[0094] The more times steps 3 and 4 are repeated, the more information grids are obtained, and the more accurate the refined remote sensing data will be. Considering the constraints of time and computing power such as the calculation cycle, in this application, the time period is generally selected to be within 5 years, and the number of corresponding atmospheric pollutant concentration remote sensing images is generally controlled to be less than 1000 frames.

[0095] Step 6: Summarize all the obtained information grids; and summarize all the atmospheric pollutant concentration information in the information grids into a single frame of atmospheric pollutant concentration remote sensing image, thereby obtaining refined remote sensing data.

[0096] Since all information grids are obtained by dividing the same grid, the process of obtaining refined remote sensing data in step 6 includes the following sub-steps:

[0097] Sub-step 6-1: Group together the information grids representing the same sub-region, that is, group together the information grids obtained by cutting the same grid unit, extract the pixels at the same position in each information grid, take the average value to obtain the new pixel, repeat the process of extracting the pixel and taking the average value multiple times, traverse each pixel in each information grid, and then obtain a new grid unit composed of the new pixels.

[0098] For example, suppose an information grid contains 100 pixels, and there are 22 information grids representing the same sub-region. Then, extract the first pixel from these 22 information grids, take their average value as the first new pixel, extract the second pixel from these 22 information grids, take their average value as the second new pixel, and so on, performing this process 100 times to obtain 100 new pixels. This results in a new information grid, which is the new grid unit. Further, the information grids representing another sub-region are grouped together to obtain corresponding new grid units, until all new grid units are obtained.

[0099] Sub-step 6-2 arranges all the new grid cells according to their actual positional relationships to obtain refined remote sensing data.

[0100] Compared to directly obtained remote sensing images of atmospheric pollutant concentrations, the refined remote sensing data obtained in step 6 largely eliminates interference caused by cloud cover and pollutant movement, providing a direct reflection of local atmospheric pollutant emissions in the monitored area. This pinpoints the emission sources of atmospheric pollutants, providing data support for subsequent emission reduction and control. In practice, analyzing local atmospheric pollution source emissions using this refined remote sensing data yields relatively accurate and practical results.

[0101] To further improve the accuracy of the refined remote sensing data, this application also provides a method for further correcting the refined remote sensing data, as follows:

[0102] Step 7: Add the grid from Step 1 and the label boxes from Step 2 to the refined remote sensing data obtained in Step 6, filter out areas with high pollution values ​​outside the label boxes as areas to be processed, and further analyze whether there are local pollution sources in the areas to be processed, and update the refined remote sensing data accordingly.

[0103] The specific selection of areas with higher pollution values ​​can be set according to actual needs. Generally, the top 20%-40% of areas with the highest pollutant concentration values ​​are selected as areas with higher pollution values.

[0104] The area to be processed includes an outline and the corresponding atmospheric pollutant concentration. The outline is a rectangular frame containing at least one pixel. The atmospheric pollutant concentration corresponding to the area to be processed is the average value of the atmospheric pollutant concentrations of the pixels it contains, denoted as the mean concentration. This means the atmospheric pollutant concentration information of the area to be processed recorded in the refined remote sensing data.

[0105] In step 7, after obtaining the area to be processed, the following sub-steps are specifically performed:

[0106] Sub-step 7-1: Retrieve the remote sensing image and its label information obtained in step 3, and add the area to be processed and the grid to the remote sensing image; retrieve the near-surface wind information of the area to be detected; the wind information is the wind information of the date corresponding to the remote sensing image and the wind information of the previous 3 days, and the wind information includes wind speed and wind direction; the near-surface wind information is obtained by retrieving information provided by the local meteorological monitoring station or local weather forecast information.

[0107] During the addition process, the outline is redrawn at the same location in the remote sensing image to complete the addition process. The pixels contained in the redrawn outline correspond to the average concentration of air pollutants, which is recorded as the observed concentration, i.e., the air pollutant concentration information of the area to be processed reflected by a frame of remote sensing image retrieved in sub-step 7-1.

[0108] Sub-step 7-2: Compare the mean concentration and the observed concentration of the area to be processed. When the observed concentration is less than half of the mean concentration, add a mark to the area to be processed.

[0109] The markers are virtual data and cannot be visually seen in remote sensing images. Their purpose is to provide clues and statistics, and to provide a data basis for subsequent processing of the area to be processed based on the number of markers.

[0110] Through the preliminary comparison in step 7-2, remote sensing images of atmospheric pollutant concentrations with significantly lower pollutant concentrations can be selected. If the proportion of such remote sensing images is high, it can be considered that the atmospheric pollutants emitted in the area to be treated are relatively few, and the remote sensing data for refined processing should be adjusted accordingly.

[0111] Sub-step 7-3: When the observed concentration is not less than half of the mean concentration, analyze whether the air pollutants emitted at the high-value label in the grid cell where the area to be treated is located can be moved to the area to be treated by the wind; if so, add a label to the area to be treated.

[0112] In this sub-step 7-3, based on the wind conditions in the area over 3 days, the distance A that the pollutant at the high-value label can move is determined, and the distance B between the center of the high-value label and the center of the area to be treated is determined. The analysis is performed by comparing the magnitudes of distance A and distance B.

[0113] This sub-step 7-3 further analyzes the potential impact of pollutant migration on the area to be treated. If there are many remote sensing images of atmospheric pollutant concentrations affected by pollutant migration, it can be determined that the high pollution values ​​reflected in the remote sensing images of atmospheric pollutant concentrations in the area to be treated are due to pollutant migration. The area to be treated basically does not have local pollution sources, and the remote sensing data for the refined processing should be adjusted accordingly.

[0114] Sub-step 7-4 involves repeating sub-steps 7-1, 7-2, and 7-3 sequentially, selecting different remote sensing images of atmospheric pollutant concentrations each time sub-step 7-1 is executed, until all remote sensing images used in step 3 have been traversed. The number of labels N for each area to be processed and the number M of times each grid cell containing that area is identified as an information grid are counted. When N is greater than 70% of M, the atmospheric pollutant concentration for that area is reassigned, indicating that the area is likely to lack a local pollution source. Conversely, if N and M do not meet the above requirements, no special processing is performed; the area is considered to have a high probability of containing a local pollution source and requires further verification and processing.

[0115] The process of reassigning atmospheric pollutant concentration is as follows: read the markers added to the area to be processed, retrieve the remote sensing images used in sub-step 7-1 corresponding to each marker, read the pixels representing the area to be processed in these remote sensing images, and obtain a new pixel by averaging the pixels at the same position. By traversing all pixels, the remote sensing images that make up the new area to be processed can be obtained, and the refined remote sensing data is updated accordingly, that is, the area to be processed in the refined remote sensing data is replaced with the new area to be processed.

[0116] In this application, the refined remote sensing data is further updated and corrected through the method in step 7, so that the local air pollutant emission information reflected in the refined remote sensing data is more accurate, and emission reduction and control work can be carried out accordingly. Example

[0117] Step 1: Select Shandong Province as the area to be detected, retrieve a high spatial resolution remote sensing image of the area, and draw a square grid with a side length of 50 km in the image. This grid divides the area to be detected into 75 sub-regions, as shown below. Figure 2 As shown;

[0118] Step 2: In the high spatial resolution remote sensing image, select high-value labels in areas where cities, industrial parks, traffic emission areas and polluting enterprises are located, select medium-value labels in areas where farmland, desert and forest are located, and select low-value labels in areas where lakes, reservoirs and seas are located.

[0119] Step 3: Retrieve the tropospheric NO2 column concentration containing the area to be detected as a remote sensing image of atmospheric pollutant concentration; add labels and a grid to the remote sensing image, and use the grid to divide the remote sensing image into 75 grid cells. Figure 3 This diagram illustrates the tropospheric NO2 column concentration distribution map of Shandong Province on May 1, 2022, divided into 75 grid cells. Further, each label is assigned a value; the label assignment process involves reading the pollutant concentration values ​​of all pixels falling within the area corresponding to the label, and taking their average value as the pollutant concentration value for that label, thus obtaining a complete label. The 20th grid cell and its labels are shown below. Figure 4 As shown in the figure; where hig represents the high value label and mid represents the median value label.

[0120] Step 4: Analyze each grid cell one by one to determine if the information in the grid cell is normal. Figure 4 In the grid cell, there are 3 high-value labels, 7 median labels, and 0 low-value labels. Further reading of the label values ​​shows that the high-value labels are all greater than any median label value and also greater than any low-value label value. The atmospheric pollutant concentration information recorded in the remote sensing image of this grid cell is saved, and this grid cell is an information grid.

[0121] Step 5: Repeat steps 3 and 4 30 times. Each time step 3 is repeated, retrieve remote sensing images of atmospheric pollutant concentrations in the area to be detected from different dates within the month. Convert each frame of remote sensing image into a set of information grids and empty grids.

[0122] Step 6: Summarize all obtained information grids; and summarize all atmospheric pollutant concentration information in the information grids into a single frame of atmospheric pollutant concentration remote sensing imagery, thereby obtaining refined remote sensing data, such as... Figure 5 As shown, the Figure 5 The pollutant concentration information shown in the image represents the local air pollution sources.

[0123] Traditional methods for identifying local air pollution sources involve directly reading data such as... Figure 3 The tropospheric NO2 column concentration shown in the figure indicates that areas with higher pollutant concentrations are local sources of air pollution.

[0124] exist Figure 3 Arbitrarily select a highly polluted area, namely the 46th grid cell. Figure 6 The middle shows Figure 3 The tropospheric NO2 column concentration distribution map of the 46th grid unit in Shandong Province on May 1, 2022, from... Figure 3 and Figure 6 It can be seen that this area is a highly polluted area with local sources of air pollution.

[0125] Retrieve accordingly Figure 5 The 46th grid cell in the middle, as Figure 7 As shown, from Figure 5 and Figure 7 This indicates that the area is not a highly polluted area and there are no local sources of air pollution.

[0126] Further retrieve the high-resolution ground feature image corresponding to the 46th grid cell, such as... Figure 8 As shown, most of the surface corresponding to the 46th grid cell is ocean, and no local air pollution sources were found in this area after field investigation. Figure 5 The local air pollution source information displayed is accurate. Figure 3 The local air pollution source information displayed is inaccurate. This demonstrates that the method for identifying local air pollution sources based on refined remote sensing data provided in this application can provide more accurate local air pollution source information.

[0127] The present invention has been described above with reference to preferred embodiments; however, these embodiments are merely exemplary and illustrative. Various substitutions and modifications can be made to the present invention based on these embodiments, all of which fall within the scope of protection of the present invention.

Claims

1. A method for identifying local atmospheric pollution sources based on remote sensing data based on refinement processing, characterized in that, The method comprises the following steps: Step 1, call high spatial resolution remote sensing image containing the region to be detected, draw a grid in the remote sensing image, the grid divides the region to be detected into a plurality of sub-regions with fixed size; Step 2, establish a label frame in the high spatial resolution remote sensing image; Step 3, call atmospheric pollutant concentration remote sensing image containing the region to be detected, add the label frame and the grid to the remote sensing image, divide the remote sensing image into a plurality of grid units with the grid, and assign a value to each label frame; Step 4, analyze each grid unit one by one to determine whether the information in the grid unit is normal, if normal, save the atmospheric pollutant concentration information recorded in the remote sensing image in the grid unit, the grid unit is an information grid; if not normal, clear the atmospheric pollutant concentration information recorded in the remote sensing image in the grid unit, the grid unit is an empty grid; Step 5, repeat steps 3 and 4, call atmospheric pollutant concentration remote sensing image containing the region to be detected at different times each time step 3 is repeated; convert each frame of remote sensing image into a set of information grids and empty grids; Step 6, aggregate all the information grids obtained; and aggregate all the atmospheric pollutant concentration information in the information grids into a high spatial resolution remote sensing image of atmospheric pollutant concentration remote sensing image, thereby obtaining refined remote sensing data.

2. The method for identifying local atmospheric pollution sources from remote sensing data based on refinement processing according to claim 1, characterized in that, The label frame comprises a high-value label frame, a medium-value label frame and a low-value label frame.

3. The method for identifying local atmospheric pollution sources based on refined remote sensing data according to claim 2, characterized in that, In the step 2, the process of establishing a label comprises the following sub-steps: Sub-step 2-1, draw the outline of the high-concentration atmospheric pollutant emission area in the high spatial resolution remote sensing image, the high-concentration atmospheric pollutant emission area includes cities, industrial parks, traffic emission areas and areas where enterprises discharge pollutants; Take the union of the outlines of the plurality of high-concentration atmospheric pollutant emission areas drawn to obtain a plurality of preliminary high-value label outlines; Sub-step 2-2, draw the outline of the low-concentration atmospheric pollutant emission area in the high spatial resolution remote sensing image, the low-concentration atmospheric pollutant emission area includes farmland, desert and forest area; Take the union of the outlines of the plurality of low-concentration atmospheric pollutant emission areas drawn to obtain a plurality of preliminary medium-value label outlines; Sub-step 2-3, draw the outline of the area without atmospheric pollutant emission in the high spatial resolution remote sensing image, the area without atmospheric pollutant emission includes lakes, reservoirs and seas; Take the union of the outlines of the plurality of areas without atmospheric pollutant emission drawn to obtain a plurality of preliminary low-value label outlines; Sub-step 2-4, select a rectangular area in the label outline, the side length of the rectangular area is an integer multiple of the minimum pixel point side length of the atmospheric pollutant concentration remote sensing image.

4. The method for identifying local atmospheric pollution sources based on refined remote sensing data according to claim 3, characterized in that, Through the above union process, each sub-region corresponds to at least one preliminary high-value label outline and at least one preliminary medium-value label outline.

5. The method of claim 3, wherein the rectangular region is completely within the label contour, and when the label contour is too small to completely cover the minimum pixel point, the label contour information is abandoned and deleted.

6. The method of claim 1, wherein the label frame assignment process in step 3 is to read the pollutant concentration values of all pixel points in the label frame in the remote sensing image, and take the average value as the pollutant concentration value of the label frame, thereby obtaining a complete label.

7. The method of claim 2, wherein in step 4, when the value of any high-value label is greater than the value of any medium-value label and the value of any low-value label, the information in the grid cell is normal, otherwise the information in the grid cell is considered abnormal.

8. The method of claim 1, wherein the process of obtaining the refined remote sensing data in step 6 includes the following sub-steps: Sub-step 6-1: The grid cells representing the same sub-region are collected together, the pixel points at the same position in each grid cell are extracted, the average value is taken to obtain a new pixel point, and the process of extracting pixel points and taking the average value is repeated multiple times to traverse each pixel point in each grid cell, thereby obtaining new grid cells composed of new pixel points; Sub-step 6-2: All new grid cells are arranged according to the real position relationship to obtain refined remote sensing data.

9. The method of claim 2, wherein the method further includes the following step 7: Step 7: Adding the grid in step 1 and the label frame in step 2 to the refined remote sensing data obtained in step 6, screening out the area with high pollution value outside the label frame as a to-be-processed area, and further analyzing whether the to-be-processed area has a local pollution source, thereby updating the refined remote sensing data.

10. The method of claim 9, wherein after obtaining the to-be-processed area in step 7, the following sub-step is performed: Sub-step 7-1: Call the remote sensing image and the label information in it obtained in step 3, add the to-be-processed area and the grid to the remote sensing image, and call the near-surface wind information of the to-be-detected area. ​ ​ ​ ​ ​ ​ Sub-step 7-2, comparing the average concentration of the to-be-processed region with the observed concentration, when the observed concentration is less than half of the average concentration, a label is added to the to-be-processed region; in the refined remote sensing data, the average concentration of the atmospheric pollutant concentration of the pixel points contained in the to-be-processed region is recorded as the average concentration; in the remote sensing image called in sub-step 7-1, the average concentration of the atmospheric pollutant concentration of the pixel points contained in the to-be-processed region is recorded as the observed concentration; Sub-step 7-3, when the observed concentration is not less than half of the average concentration, it is analyzed whether the atmospheric pollutants emitted by the high-value label in the grid unit where the to-be-processed region is located can move to the to-be-processed region with the wind; if so, a label is added to the to-be-processed region; Sub-step 7-4, the above sub-steps 7-1, 7-2 and 7-3 are repeatedly executed, and different atmospheric pollutant concentration remote sensing images are selected each time sub-step 7-1 is executed until all the remote sensing images used in step 3 are traversed; the number N of labels of each to-be-processed region and the number M of times that the grid unit where each to-be-processed region is located is identified as an information grid are counted, and when the value of N is greater than 70% of the value of M, the to-be-processed region is revalued with the atmospheric pollutant concentration.

11. The method for identifying local atmospheric pollution sources from remote sensing data based on refinement processing according to claim 10, characterized in that, The revalued atmospheric pollutant concentration processing process is: reading the labels added to the to-be-processed region, calling the remote sensing images used in sub-step 7-1 corresponding to each label, reading the atmospheric pollutant concentration information of the to-be-processed region in these remote sensing images, taking the average value as the new atmospheric pollutant concentration of the to-be-processed region, and updating the refined remote sensing data according to the new atmospheric pollutant concentration.

Images