Large mountain photovoltaic environment fire danger identification method based on meteorological satellite

Abstract

The invention provides a large mountain photovoltaic environment fire danger identification method based on a meteorological satellite, and the method comprises the steps: carrying out the radiometric calibration processing based on the Himawari-8 data of a detection region, obtaining the spectral reflectivity or spectral radiation brightness of each wave band of the detection region, and obtaining a cloud shielding pixel through a cloud detection method; a normalized brightness temperature time sequence change function is established based on the time sequence change of the background brightness temperature, a brightness temperature estimation value is estimated, the maximum assumed threshold value of each stage is determined according to the brightness temperature time sequence change function, and then a time sequence method is used for distinguishing fire point pixels. And finally, carrying out credibility grading on the extracted fire point pixels to obtain a fire point grading graph, and sending out corresponding early warning according to fire point grades.

Description

Technical Field

[0001] The invention relates to the technical field of mountain fire point identification, and in particular to a large-scale mountain photovoltaic environment fire risk identification method based on meteorological satellites. Background Art

[0002] Centralized photovoltaic power generation, as an important component of clean energy, is an important development direction for all provinces, cities and major power generation groups. The continuous high temperature of wildfires keeps the insulation strength of the air gap under the line at a low level, which will make it difficult for the reclosing to succeed. Therefore, there is an urgent need for an effective method for identifying fire risks in mountain photovoltaic environments to identify fire risks in the entire demonstration site. Satellite remote sensing technology has the advantages of wide monitoring range, high spatial resolution and high sensitivity, and is widely used in surface fire monitoring.

[0003] Flannigan et al. proposed the absolute threshold method, which achieved good monitoring results by comparing the brightness temperature of the pixel channel to be judged with the set threshold. However, environmental differences in different regions will affect the accuracy of fire point identification using the absolute threshold algorithm. At present, most satellite remote sensing fire point identification uses the context method, which compares the fire point to be judged with the information of the surrounding adjacent areas, and determines it as a fire point when the difference is greater than a certain set threshold. The context method is simple to use and improves the universality of the algorithm to a certain extent, but it has limitations in sensitivity. Under solar radiation, the temperature rise in areas with sparse vegetation such as bare land and sandy land is large, which can easily lead to misjudgment of fire points. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a method for identifying fire risks in large mountain photovoltaic environments based on meteorological satellites. In order to solve the above technical problems, the technical solution adopted by the present invention is: a large-scale mountain photovoltaic environment fire risk identification method based on meteorological satellites, comprising the following steps: Step 1: Collect data. Step 2: Data classification; Step 3: Radiometric calibration of Himawari-8 remote sensing images; Step 4: Use cloud detection method to obtain cloud-occluded pixels; Step 5: Use the time series change of background brightness temperature to establish a normalized brightness temperature time series change function, and calculate the brightness temperature estimate of the cloud-blocked pixel; Step 6: Determine the maximum hypothetical threshold of each stage using the background brightness temperature time series variation curve; Step 7: Use the time series method to identify fire point pixels and perform credibility classification and corresponding warning on the extracted fire point pixels.

[0005] Preferably, in step 1, the collected data include raw data from the meteorological satellite Himawari-8, Sentinel-2 high-resolution remote sensing images, ASTER 30-meter resolution DEM data, and boundary data of the study area.

[0006] Preferably, in step 2, the high-resolution remote sensing image is first segmented at multiple scales based on spectral information, and then the segmented objects are classified by decision tree based on texture features and vegetation index to achieve automatic classification of land cover types.

[0007] Preferably, in step 3, the raw data of the meteorological satellite Himawari-8 is subjected to radiation calibration processing, a brightness temperature calibration method is used for the infrared band, and a reflectance calibration method is used for the visible light and near-infrared bands; the spectral reflectance or spectral radiation brightness of the ground object is calculated, and the brightness grayscale value of the image is converted into absolute radiation brightness, thereby realizing comparative analysis of images acquired at different times and by different sensors.

[0008] Preferably, in order to obtain cloud-blocked pixels in step 4, the original data of the meteorological satellite Himawari-8 is firstly cropped using the boundary of the study area, and then the reflectivity of channels 3 and 4 and the brightness temperature value of channel 15 are selected, and cloud pixel detection is performed using a cloud detection method, and finally the area that meets the definition of the cloud area is selected to obtain cloud-blocked pixels. Channels 3 and 4 of the Himawari-8 image data are more sensitive to clouds, and channel 15 is more sensitive to cloud top temperature. Therefore, these three channels are selected to detect clouds, which can eliminate the interference of clouds in mountain fire monitoring.

[0009] Preferably, the cloud zone definition formula is as follows: ; in, , are the reflectances of the 3rd and 4th channels respectively; is the brightness temperature value of channel 15.

[0010] Preferably, in step 5, the time series variation of the background brightness temperature is used to establish a normalized brightness temperature time series variation function, which is used for inversion of missing data for cloud detection and determination of brightness temperature thresholds.

[0011] Preferably, for the cloud-blocked pixels obtained by cloud detection, their estimated values ​​at time t are calculated based on the established normalized brightness temperature time series variation function.

[0012] Preferably, the selection of the maximum hypothesized threshold value of each stage in step 6 is performed by determining the brightness temperature change rate in each interval through the normalized brightness temperature time series change function established in step 4, and then performing classification statistics to finally determine the maximum hypothesized threshold value of each stage.

[0013] Preferably, in step 7, in order to determine the fire point pixels, the brightness temperature changes of the pixels in the study area are first calculated, and then the existing fire points, potential fire points and continuous fire points are identified through the time series method to obtain all the fire points, and the feasibility level is reassigned to the updated results to obtain the final fire point classification map, and finally the corresponding warning is issued according to the fire point level.

[0014] The present invention provides a large-scale mountain photovoltaic environment fire risk identification method based on meteorological satellites, which has the following beneficial effects: 1) The present invention uses remote sensing images of the new generation geostationary meteorological satellite Himawari-8, which has the characteristics of high frequency observation and wide coverage area. In the judgment process, cloud pixel identification is performed before fire pixel identification using Himawari-8 satellite channel observation data, and the fire point identification conditions are dynamically adjusted through cloud pixel information.

[0015] 2) The present invention adopts a time-series detection method for fire point identification, and based on the difference in pixel brightness temperature at the observation time, it can more efficiently and accurately monitor fire conditions through satellite remote sensing. The fire point identification threshold of the time-series method is significantly lower than that of the previous fire point identification method, and the corresponding sub-pixel fire point area is reduced by more than half, thereby improving the sensitivity of fire identification and realizing early detection of fire points.

[0016] 3) Improve the accuracy of fire risk identification. By collecting multi-source data, including raw data from the meteorological satellite Himawari-8, Sentinel-2 high-resolution remote sensing images, ASTER 30-meter resolution DEM data, and study area boundary data, information is obtained from different dimensions. Comprehensive analysis helps to identify fire risks more accurately. When classifying data, spectral information, texture features, and vegetation indices are used for multi-scale segmentation and decision tree classification to achieve automatic classification of land cover types, providing a basis for subsequent accurate judgment of fire risk conditions. The time series method is used to identify existing fire points, potential fire points, and continuous fire points, and the credibility of fire point pixels is graded. Compared with traditional methods, the fire risk status can be determined more carefully and accurately.

[0017] 4) The data processing is scientific and effective. The Himawari-8 remote sensing images are calibrated by radiation. The brightness temperature calibration and reflectivity calibration methods are used for different bands to convert the image brightness grayscale value into absolute radiation brightness, which is convenient for comparative analysis of images acquired at different times and by different sensors, ensuring the consistency and comparability of the data and providing reliable data support for subsequent fire risk identification. The normalized brightness temperature time series change function is established by using the background brightness temperature time series change, which can not only be used for cloud detection missing data inversion, but also determine the brightness temperature threshold and calculate the brightness temperature estimation value of cloud-blocked pixels, effectively handling the problem of missing data and improving data integrity and availability.

[0018] 5) Adapting to complex mountain photovoltaic environments. Considering the complexity of mountain photovoltaic environments, this method uses the boundaries of the study area to clip the data, and selects the reflectivity and brightness temperature values ​​of specific channels for cloud pixel detection. According to different cloud area definition formulas during the day and at night, cloud-blocked pixels are obtained, which can effectively eliminate cloud interference and accurately identify fire risks in mountain photovoltaic environments. By determining the maximum hypothetical threshold of each stage and combining the time series method to identify fire points, it can better adapt to the needs of fire risk identification under different environmental conditions, reduce misjudgments caused by environmental differences, and improve applicability in large mountain photovoltaic environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The present invention will be further described below in conjunction with the accompanying drawings and embodiments: Figure 1 is a flow chart of an embodiment of the present invention; Figure 2 This is a flow chart of fire risk remote sensing monitoring and early warning of the present invention; Figure 3 This is a flow chart of fire point identification using the sequential method of the present invention. DETAILED DESCRIPTION

[0020] like Figure 1-2 As shown, the fire risk identification method for large mountain photovoltaic environment based on meteorological satellite includes the following steps: Step 1: Collect data. Step 2: Data classification; Step 3: Radiometric calibration of Himawari-8 remote sensing images; Step 4: Use cloud detection method to obtain cloud-occluded pixels; Step 5: Use the time series change of background brightness temperature to establish a normalized brightness temperature time series change function, and calculate the brightness temperature estimate of the cloud-blocked pixel; Step 6: Determine the maximum hypothetical threshold of each stage using the background brightness temperature time series variation curve; Step 7: Use the time series method to identify fire point pixels and perform credibility classification and corresponding warning on the extracted fire point pixels.

[0021] It provides a complete set of fire risk identification processes for large-scale mountain photovoltaic environments, covering multiple links such as data collection, classification, image processing, cloud pixel detection, brightness temperature estimation, threshold determination, fire point identification and early warning. Each step is closely linked, providing a comprehensive and systematic method framework for fire risk identification, ensuring the integrity and accuracy of the identification process.

[0022] Preferably, in step 1, the collected data include raw data from the meteorological satellite Himawari-8, Sentinel-2 high-resolution remote sensing images, ASTER 30-meter resolution DEM data, and boundary data of the study area. Collect multi-source data, the raw data from the meteorological satellite Himawari-8 is used to obtain macroscopic meteorological and surface information; Sentinel-2 high-resolution remote sensing images provide more detailed surface features; ASTER 30-meter resolution DEM data can reflect topography; and the boundary data of the study area clearly defines the monitoring scope. Multi-source data complement each other, improve the accuracy and reliability of fire risk identification, and provide a rich and comprehensive information basis for subsequent analysis.

[0023] Preferably, in step 2, the high-resolution remote sensing image is first segmented at multiple scales based on spectral information, and then the segmented objects are classified by decision tree based on texture features and vegetation index, so as to realize automatic classification of land cover types. The combination of multi-scale segmentation and decision tree classification, first segmentation based on spectral information, and then classification based on texture features and vegetation index, can more accurately classify land cover types automatically. This helps to distinguish different surface covers, such as vegetation, bare land, water bodies, etc., and provides an important basis for accurately judging the environment and possibility of fire risk, thereby improving the pertinence and accuracy of fire risk identification.

[0024] Preferably, in step 3, the raw data of the meteorological satellite Himawari-8 is subjected to radiation calibration processing, the brightness temperature calibration method is used for the infrared band, and the reflectance calibration method is used for the visible light and near-infrared bands; the spectral reflectance or spectral radiation brightness of the ground object is calculated, and the brightness grayscale value of the image is converted into absolute radiation brightness, so as to achieve comparative analysis of images acquired at different times and by different sensors. Different calibration methods are used for different bands, and the image brightness grayscale value is converted into absolute radiation brightness, so that images acquired at different times and by different sensors can be compared and analyzed. This ensures the consistency and comparability of the data, provides a stable and reliable data basis for subsequent fire risk identification and analysis, and improves the credibility of the identification results.

[0025] Preferably, in order to obtain cloud-blocked pixels in step 4, the original data of the meteorological satellite Himawari-8 is firstly cropped using the boundary of the study area, and then the reflectivity of channels 3 and 4 and the brightness temperature value of channel 15 are selected, and cloud pixel detection is performed using a cloud detection method, and finally the area that meets the definition of the cloud area is selected to obtain cloud-blocked pixels. Cloud pixel detection is performed by boundary cropping and specific channel data selection, and the boundary of the study area and related channel information are effectively utilized, and cloud-blocked pixels are accurately obtained in combination with cloud detection methods. This can eliminate the interference of clouds on fire risk identification, improve the accuracy of fire risk identification, and avoid misjudgment or missed judgment due to cloud obstruction.

[0026] Preferably, the cloud zone definition formula is as follows: ; in, , are the reflectances of the 3rd and 4th channels respectively; is the brightness temperature value of channel 15. The formulas for defining different cloud areas during the day and at night fully consider the impact of the difference in day and night illumination and temperature on cloud characteristics. This differentiated definition can more accurately identify cloud areas at different times, and then more accurately obtain cloud-blocking pixels, improve the accuracy of cloud detection, and provide more reliable data support for subsequent fire risk identification.

[0027] Preferably, in step 5, the time series change of background brightness temperature is used to establish a normalized brightness temperature time series change function for inversion of missing data for cloud detection and determination of brightness temperature threshold. The function established by using the time series change of background brightness temperature can be used for inversion of missing data for cloud detection and determination of brightness temperature threshold. This solves the problem of missing data in the cloud detection process, improves the integrity of the data, and provides a scientific basis for determining the brightness temperature threshold, further improving the accuracy and reliability of fire risk identification.

[0028] Preferably, for the cloud-blocked pixels obtained by cloud detection, their estimated values ​​at time t are calculated according to the established normalized brightness temperature time series variation function. The estimated brightness temperature of the cloud-blocked pixels is calculated according to the established function, which supplements the data information of the cloud-blocked area, so that the temperature conditions of the area can be reasonably estimated even in the case of cloud blockage, providing more abundant data for comprehensive and accurate analysis of fire risk conditions, and reducing the impact of information loss caused by cloud blockage on fire risk identification.

[0029] Preferably, the selection of the maximum hypothetical threshold value of each stage in step 6 is to determine the brightness temperature change rate in each interval through the normalized brightness temperature time series change function established in step 4, and then classify and count to finally determine the maximum hypothetical threshold value of each stage. This method can scientifically and reasonably determine the threshold value according to the actual brightness temperature change situation by determining the brightness temperature change rate based on the brightness temperature time series change function and classify and count to select the maximum hypothetical threshold value, so as to improve the accuracy and adaptability of the threshold value, provide a more reliable standard for the identification of fire point pixels, and thus improve the accuracy of fire risk identification.

[0030] Preferably, in order to determine the fire point pixel in step 7, the pixel brightness temperature change in the study area is first calculated, and then the existing fire points, potential fire points and continuous fire points are identified by the time series method fire point, and all fire points are obtained, and the updated results are re-assigned with a feasibility level to obtain a final fire point classification map, and finally a corresponding warning is issued according to the fire point level. By calculating the pixel brightness temperature change and the time series method fire point identification, various types of fire points can be fully identified, and the fire points can be graded and warned for credibility. This can not only detect fire points in time, but also evaluate the degree of danger of fire points, so as to take appropriate measures to effectively ensure the safety of mountain photovoltaic environment.

[0031] Figure 3 For the specific operation of fire point identification using the time series method, the brightness temperature data of all pixels in the study area at different observation times are first processed to calculate the change of pixel brightness temperature over time. Then, the existing fire points are identified based on the calculated pixel brightness temperature change data combined with the determined threshold value. Then, the potential fire point judgment factor is calculated and combined with the discrimination conditions to find the potential fire point. Then, the continuous fire point judgment factor is calculated and combined with continuous observation to determine the continuous fire point, and then various types of fire points are integrated. The integrated fire points are re-evaluated based on the brightness temperature change amplitude, duration, surrounding environment and other information, and each fire point is assigned a corresponding credibility level. Then, based on the re-assigned credibility level, the fire points in the study area are visualized on the map to generate a fire point classification map. Finally, according to the fire point level and the pre-set warning rules, the corresponding level of warning information is issued to ensure the safety of the mountain photovoltaic environment through timely and accurate warnings.

[0032] The present invention provides a large-scale mountain photovoltaic environment fire risk identification method based on meteorological satellite. Firstly, radiation calibration processing is performed based on Himawari-8 data of the measurement area to obtain the spectral reflectance or spectral radiation brightness of each band in the measurement area, and then a cloud detection method is used to obtain cloud-blocked pixels. Based on the time series change of background brightness temperature, a normalized brightness temperature time series change function is established and the brightness temperature estimation value is estimated. According to the brightness temperature time series change function, the maximum assumption threshold of each stage is determined, and then the time series method is used to distinguish the fire point pixels. Finally, the extracted fire point pixels are credibility graded to obtain a fire point classification map and a corresponding early warning is issued according to the fire point level.

[0033] The above embodiments are only preferred technical solutions of the present invention and should not be regarded as limiting the present invention. The protection scope of the present invention shall be the technical solutions recorded in the claims, including equivalent replacement solutions of the technical features in the technical solutions recorded in the claims. That is, equivalent replacement improvements within this scope are also within the protection scope of the present invention.

Claims

1. A large-scale mountain photovoltaic environment fire risk identification method based on meteorological satellites, characterized in that: The following steps are involved: Step 1: Collect data. Step 2: Data classification; Step 3: Radiometric calibration of Himawari-8 remote sensing images; Step 4: Use cloud detection method to obtain cloud-occluded pixels; Step 5: Use the time series change of background brightness temperature to establish a normalized brightness temperature time series change function, and calculate the brightness temperature estimate of the cloud-blocked pixel; Step 6: Determine the maximum hypothetical threshold of each stage using the background brightness temperature time series variation curve; Step 7: Use the time series method to identify fire point pixels and perform credibility classification and corresponding warning on the extracted fire point pixels.

2. The large-scale mountain photovoltaic environment fire risk identification method based on meteorological satellite according to claim 1 is characterized in that: In step 1, the collected data include raw data from the meteorological satellite Himawari-8, high-resolution remote sensing images from Sentinel-2, 30-meter resolution DEM data from ASTER, and boundary data of the study area.

3. The large-scale mountain photovoltaic environment fire risk identification method based on meteorological satellite according to claim 2 is characterized in that: The step 2 first performs multi-scale segmentation on the high-resolution remote sensing image based on spectral information, and then performs decision tree classification on the segmented objects based on texture features and vegetation index to achieve automatic classification of land cover types.

4. The large-scale mountain photovoltaic environment fire risk identification method based on meteorological satellite according to claim 1 is characterized in that: In the step 3, the raw data of the meteorological satellite Himawari-8 is subjected to radiation calibration processing, a brightness temperature calibration method is used for the infrared band, and a reflectance calibration method is used for the visible light and near-infrared bands; the spectral reflectance or spectral radiation brightness of the ground object is calculated, and the brightness grayscale value of the image is converted into absolute radiation brightness, so as to realize comparative analysis of images acquired at different times and by different sensors.

5. The large-scale mountain photovoltaic environment fire risk identification method based on meteorological satellite according to claim 1 is characterized in that: In order to obtain cloud-blocked pixels in step 4, the original data of the meteorological satellite Himawari-8 is first cropped using the boundary of the study area, and then the reflectivity of channels 3 and 4 and the brightness temperature value of channel 15 are selected, and cloud pixels are detected using a cloud detection method. Finally, the area that meets the definition of the cloud area is selected to obtain cloud-blocked pixels.

6. The large-scale mountain photovoltaic environment fire risk identification method based on meteorological satellite according to claim 5 is characterized in that: The cloud zone definition formula is as follows: ； in, , are the reflectances of the 3rd and 4th channels respectively; is the brightness temperature value of channel 15.

7. The large-scale mountain photovoltaic environment fire risk identification method based on meteorological satellite according to claim 1 is characterized in that: In step 5, the time series variation of the background brightness temperature is used to establish a normalized brightness temperature time series variation function, which is used for the inversion of missing data for cloud detection and the determination of the brightness temperature threshold.

8. The method for identifying fire risk in a large mountain photovoltaic environment based on a meteorological satellite according to claim 7 is characterized in that: Then, for the cloud-blocked pixels obtained by cloud detection, their estimated values ​​at time t will be calculated based on the established normalized brightness temperature time series change function.

9. The large-scale mountain photovoltaic environment fire risk identification method based on meteorological satellite according to claim 1 is characterized in that: The selection of the maximum hypothetical threshold of each stage in step 6 is to determine the brightness temperature change rate in each interval through the normalized brightness temperature time series change function established in step 4, and then perform classification statistics to finally determine the maximum hypothetical threshold of each stage.

10. The large-scale mountain photovoltaic environment fire risk identification method based on meteorological satellite according to claim 1 is characterized in that: In step 7, in order to determine the fire point pixels, first calculate the brightness temperature changes of the pixels in the study area, then identify the existing fire points, potential fire points and continuous fire points through the time series method, obtain all the fire points, and re-assign the feasibility level of the updated results to obtain the final fire point classification map, and finally issue corresponding warnings according to the fire point level.

Images