A spatiotemporal adaptive forest fire monitoring method and verification method based on FY3C data

Abstract

The invention discloses a spatiotemporal adaptive forest fire monitoring method and a testing method based on FY3C data, comprising the following steps: step 1: obtaining required data and preprocessing the data; step 2: identifying a forest fire concern area according to a binary method based on the preprocessed data; step 3: identifying cloud pixels from the forest fire concern area obtained in step 2; step 4: performing forest fire monitoring and identification on non-cloud pixels in the forest fire concern area in step 2; the identification process is as follows: S1: extracting abnormally high temperature points as potential fire points from non-cloud pixels in the forest fire concern area; S2: extracting absolute fire points from potential fire points according to the brightness temperature of a third channel; S3: extracting possible fire points other than absolute fire points from potential fire points according to set discrimination conditions; the invention identifies hot spots by a threshold method, determines the threshold value according to changes in region, time and underlying surface, and avoids defects caused by excessively high or low threshold values.

Description

Technical Field

[0001] The present invention relates to the technical field of forest fire, and in particular to a spatiotemporal adaptive forest fire monitoring method and a testing method based on FY3C data. Background Art

[0002] Forests are an important natural resource and play an important role in the maintenance and restoration of the ecological environment. Forest fires directly lead to the reduction of forest resource area, ecological environmental degradation, and serious threats to the diversity of surface vegetation and wild species. Timely and accurate detection of forest fires is of great significance to the protection of forest resources, the maintenance of the ecological environment, the protection of species diversity, and the regulation of climate. Remote sensing data is widely used in forest fire monitoring due to its wide coverage and high temporal and spatial resolution. However, the current monitoring methods are relatively cumbersome to operate, and a large number of forest fires are still missed and falsely reported during the monitoring process. Summary of the invention

[0003] The present invention aims at solving the problems in the prior art and provides a spatiotemporal adaptive forest fire monitoring method and a testing method based on FY3C data.

[0004] The technical solution adopted by the present invention is:

[0005] A spatiotemporal adaptive forest fire monitoring method based on FY3C data comprises the following steps:

[0006] Step 1: Obtain the required data and preprocess the data;

[0007] Step 2: Identify forest fire concern areas based on the preprocessed data using the binary method;

[0008] Step 3: Identify cloud pixels from the forest fire concern area obtained in step 2;

[0009] Step 4: Perform forest fire monitoring and identification on the non-cloud pixels in the forest fire focus area in step 2;

[0010] The identification process is as follows:

[0011] S1: Extract abnormally high temperature points from non-cloud pixels in the forest fire concern area as potential fire points;

[0012] S2: Extract the absolute fire point from the potential fire points according to the brightness temperature of the third channel;

[0013] S3: Extract possible fire points other than absolute fire points from potential fire points according to the set discrimination conditions.

[0014] Furthermore, the method for identifying the forest fire concern area in step 2 is as follows:

[0015] Identify non-forest fire concern areas and mark them with a value of 1, and the rest of the areas are forest fire concern areas and are marked with a value of 0; use the forest fire concern area and non-concern area binary layers as masks to extract the forest fire concern areas used to detect the image;

[0016] Among them, the non-forest fire concern area extraction method extracts urban and rural construction land, road land, industrial and mining land, paddy fields, and fixed water bodies based on surface cover data; detects variable water bodies based on remote sensing images; variable water bodies meet the following conditions:

[0017] The pixel satisfies R2<α1 and

[0018] Wherein, R1 is the reflectivity of the first channel, R2 is the reflectivity of the second channel, and α1 and α2 are set thresholds.

[0019] Furthermore, the method for identifying cloud pixels in step 3 is as follows:

[0020] If a pixel meets the following conditions during the day, it is a cloud pixel:

[0021] R1>α3 and T5>α4

[0022] Or meet the following conditions:

[0023] R1+R2>α5 and

[0024] At the same time, the pixels satisfy the following conditions at night:

[0025] T5>α7

[0026] Among them, R1 is the reflectivity of the first channel, R2 is the reflectivity of the second channel, T5 is the brightness temperature of the fifth channel, and α3, α4, α5, α6, and α7 are set thresholds.

[0027] Furthermore, in step 4, before forest fire identification, forest fire monitoring units and monitoring sub-areas are first divided; a brightness temperature query library is established according to the forest fire monitoring units and monitoring sub-areas, and then forest fire identification is performed;

[0028] The division method of forest fire monitoring units and monitoring sub-areas is as follows:

[0029] The data is partitioned according to the latitude and longitude range and terrain. Each scanned image is divided into an independent monitoring unit according to the range of 1°×1°, and the monitoring sub-area is divided according to the elevation setting threshold within the monitoring unit;

[0030] The method for establishing the brightness temperature query library is as follows:

[0031] The 4th channel is used to calculate and build tables for the long-term multi-period monitoring surface brightness temperature values ​​of the forest fire monitoring unit and the monitoring sub-area; the long-term series duration is set according to needs, and the period length is set to one quarter or one month. The mean daytime brightness temperature and the mean nighttime brightness temperature are calculated within the period; based on the calculated data, a brightness temperature query database for different monitoring sub-areas in the monitoring unit is established.

[0032] Furthermore, the method for extracting abnormally high temperature points in step S1 is as follows:

[0033] The non-cloud pixels in the forest fire focus area meet the requirement that the brightness temperature of the third channel is greater than the set threshold T h , the brightness temperature of the 4th channel is greater than the value of the brightness temperature query library for this period, and the reflectivity of the 1st and 2nd channels is less than the set threshold at the same time.

[0034] Furthermore, the method for extracting the absolute fire point according to the brightness temperature of the third channel in step S2 is as follows:

[0035] For non-cloud pixels in the forest fire concern area, the brightness temperature of the third channel must be greater than the set threshold during the day and the threshold must be set at night.

[0036] Furthermore, the specific method of step S3 is as follows:

[0037] Non-cloud pixels in the forest fire concern area and non-absolute fire point pixels meet A or B and C:

[0038] A: T3-T 3b >4δT 3b

[0039] B: T3-T 3b >ΔT3

[0040] C: T3-T4-T 34b >4δT 34b

[0041] Among them, T3 is the brightness temperature value of the 3rd channel, T4 is the brightness temperature value of the 4th channel, T 3b is the average brightness temperature of the background pixels in channel 3, T 34b is the background temperature mean of the brightness temperature difference between channels 3 and 4, δT 3b is the standard deviation of the brightness temperature of the background pixels in channel 3, δT 34b is the background temperature standard deviation of the brightness temperature difference between channels 3 and 4, and ΔT3 is the brightness temperature anomaly of channel 3.

[0042] A verification method for a spatiotemporal adaptive forest fire monitoring method based on FY3C data was proposed, which compared the identified potential fire points with the real fire points for verification, including the verification of forest fire location and forest fire area.

[0043] Further, the forest fire location inspection method is as follows:

[0044] If the monitored forest fire point falls within the real forest fire pixel or at a set position around the real forest fire pixel, the monitored forest fire point is consistent with the actual forest fire point.

[0045] Furthermore, the forest fire area inspection method is as follows:

[0046] Open fire area S Fde and the actual forest fire area S Fac If the area ratio is within the set threshold, the monitored forest fire area is considered to have passed the inspection;

[0047] S Fde =P×S

[0048] Among them, P is the percentage of the open fire area to the pixel area, and S is the area of ​​a single pixel;

[0049] P can be obtained as follows:

[0050]

[0051] Among them, N if is the mixed pixel radiance, N ihi is the emissivity of the open fire area, N ibg is the radiation rate of the background area around the fire point, T hi is the temperature of the open flame zone, C1 and C2 are constants, T bg is the background temperature;

[0052] P and T hi Solved by Newton's iteration method.

[0053] The beneficial effects of the present invention are:

[0054] (1) The present invention identifies hot spots through a threshold method, and determines the threshold value according to the changes in region, time and underlying surface, thus avoiding the defects caused by a threshold that is too high or too low;

[0055] (2) The present invention utilizes surface cover data to resolve the interference of other ground heat sources on forest fire monitoring, and can effectively identify forest fire hotspots and accurately distinguish fire points. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 Schematic diagram of a forest fire location detection method in an embodiment of the present invention.

[0057] Figure 2 It is a schematic diagram of the monitoring method flow of the present invention.

[0058] Figure 3 These are the original images of the 3rd and 4th channels in the embodiment of the present invention.

[0059] Figure 4 The figure is a schematic diagram of the results of monitoring forest fire points using the method of the present invention and the traditional method. DETAILED DESCRIPTION

[0060] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

[0061] like Figure 2 As shown, a spatiotemporal adaptive forest fire monitoring method based on FY3C data includes the following steps:

[0062] Step 1: Get the required data and preprocess the data; preprocessing includes radiation calibration, atmospheric correction, and geometric correction; the purpose of preprocessing is to convert the original DN value of the remote sensing image into the reflectivity of the outer surface of the atmosphere (i.e., radiation brightness) through radiation calibration. The surface reflectivity or radiation brightness is converted into the actual reflectivity (i.e., radiation brightness) of the ground surface through atmospheric correction to eliminate the errors caused by atmospheric scattering, absorption, and reflection, and then the extrusion, stretching, distortion, or offset of the pixel relative to the actual position of the ground caused by systematic and non-systematic factors is corrected through geometric correction to achieve the purpose of matching the pixel with the actual position. After preprocessing, it is also necessary to go through projection, format conversion, and data clipping; projection is to match the remote sensing data with the projection method you need, format conversion is to convert it into a data format that you can process, and data clipping is to obtain local data of research concern and narrow the area of ​​concern.

[0063] Step 2: Identify forest fire concern areas based on the preprocessed data using the binary method;

[0064] The surface cover data closest to the time period of remote sensing data acquisition was obtained, and urban and rural construction land, road land, industrial and mining land, paddy fields, and fixed water bodies were extracted to generate a binary map. The urban and rural construction land was assigned a value of 1, and the remaining pixels were assigned a value of 0.

[0065] Detect variable water bodies based on remote sensing images; variable water bodies meet the following conditions:

[0066] The pixel satisfies R2<α1 and

[0067] Among them, R1 is the reflectance of the first channel, R2 is the reflectance of the second channel, and α1 and α2 are set thresholds. α1 is 0.1 and α2 is 0.05. The binary map layer is superimposed on this layer to extract non-water pixels and pixels with a binary map value of 0, which are defined as forest fire concern areas. Forest fire concern areas include dry fields, woodlands (including gardens, fruit trees and other types of woodlands), grasslands and other areas with lush vegetation.

[0068] In addition to the above methods, forest fire concern areas can also be identified through neural networks.

[0069] The following steps are involved:

[0070] First, build a neural network. You can choose a residual neural network.

[0071] Process the data obtained in step 1 (in order to meet the input requirements of the neural network) and divide it into a training set and a test set;

[0072] Then the training set is used to input the neural network, the neural network is trained, and the structural parameters of the neural network are determined;

[0073] The preprocessed data is input into the trained neural network to obtain the recognition result. The neural network method can be combined with the above binary method to identify the forest fire concern area.

[0074] Step 3: Identify cloud pixels from the forest fire concern area obtained in step 2;

[0075] The method for identifying cloud pixels is as follows:

[0076] If a pixel meets the following conditions during the day, it is a cloud pixel:

[0077] R1>α3 and T5>α4

[0078] Or meet the following conditions:

[0079] R1+R2>α5 and

[0080] At the same time, the pixels satisfy the following conditions at night:

[0081] T5>α7

[0082] Among them, R1 is the reflectivity of the first channel, R2 is the reflectivity of the second channel, T5 is the brightness temperature of the fifth channel, α3 is 0.1, α4 is 289, α5 is 0.3, α6 is 0.3, and α7 is 285.

[0083] Step 4: Monitor and identify forest fires in non-cloud pixels in the forest fire concern area;

[0084] Before forest fire identification, forest fire monitoring units and monitoring sub-areas are first divided; a brightness temperature query library is established based on the forest fire monitoring units and monitoring sub-areas, and then forest fire identification is performed;

[0085] The division method of forest fire monitoring units and monitoring sub-areas is as follows:

[0086] The data is partitioned according to the latitude and longitude range and terrain. Each scanned image is divided into an independent monitoring unit according to the range of 1°×1°, and the monitoring sub-area is divided according to the elevation setting threshold within the monitoring unit;

[0087] The height of the grid point in the monitoring unit is H i , the average elevation is H mean , calculate the difference between the grid point elevation and the average elevation as H i -H mean , H i -H mean ≥400, H i -H mean ≤-400, 400<H i -H mean Areas with elevation differences less than -400 are divided into sub-monitoring areas. If the elevation difference is between ±400, the monitoring unit is no longer divided.

[0088] The monitoring unit is defined by the coordinates of the upper left corner. The longitude and latitude of the upper left corner of the single-view scan data without any cutting or cropping are L x , L y , let the monitoring unit be S, then any monitoring unit is S i , i represents the unit position. The position of any unit is represented by S x+i,y-i , where i is a positive integer from 0 to n, x and y are the longitude and latitude values ​​of the upper left corner, respectively, and the sub-area table of each unit is regarded as S Hr , where H is the height and r is the numerical range of the height.

[0089] The method for establishing the brightness temperature query library is as follows:

[0090] The 4th channel is used to calculate and build tables for the long-term multi-period monitoring surface brightness temperature values ​​of the forest fire monitoring unit and the monitoring sub-area; the long-term series duration is set as needed (usually at least 3 years of data), and the period length is set to one quarter or one month. The mean daytime brightness temperature and the mean nighttime brightness temperature are calculated within the period; a brightness temperature query database for different monitoring sub-areas in the monitoring unit is established based on the calculated data.

[0091] Assume monitoring unit S x+i,y-i The monitoring sub-area is The average brightness temperature of any monitoring sub-area during the day is Td m The average brightness temperature at night is Tn m When calculating the quarterly mean, m is a positive integer from 1 to 4, m = 1 for spring, m = 2 for summer, m = 3 for autumn, and m = 4 for winter; when calculating the monthly mean, m is the month from 1 to 12. In this way, a brightness temperature query database is established in different monitoring sub-areas of each monitoring unit.

[0092] The identification process includes:

[0093] S1: Extract abnormally high temperature points from non-cloud pixels in the forest fire concern area as potential fire points;

[0094] The non-cloud pixels in the forest fire focus area meet the requirement that the brightness temperature of the third channel is greater than the set threshold T h , and the reflectivity of the 1st and 2nd channels is less than the set threshold at the same time.

[0095] Pixels should meet the following requirements: During the day, pixels should meet T3>T h , and the first channel reflectivity ρ 0.65 <0.3, reflectivity of the second channel ρ 0.86 <0.3; at night, T4>T h , where T3 is the brightness temperature of the third channel.

[0096] T h The calculation method is as follows:

[0097] The first method is to first calculate the cumulative probability density of the 4um channel for non-cloud pixels in the forest fire concern area according to the surface classification (forest, grassland, dry land), and arrange them from high to low according to brightness temperature. The corresponding T4 when the cumulative probability density is 8% is used as the threshold T h .

[0098] The second method is to use the average brightness temperature of different land types in long time series as the threshold Th. Taking Chongqing as an example, the reference values ​​of ΔT3 during the day and at night are shown in the following table:

[0099] Table 1. ΔT3 values ​​in different seasons

[0100]

[0101]

[0102] In addition to the above methods, the following methods can also be used to extract potential fire points:

[0103] First, a sensor network is deployed to collect weather condition parameters such as temperature, relative humidity, rainfall and wind speed measured by sensors; the sensor network is deployed according to the region, and based on the data measured by the sensor network and the above-mentioned threshold judgment method, it is determined whether it is a potential fire point.

[0104] The following formula is used to express the threat level of fire:

[0105] Y=β1T j +β2W fj +β3P dj +β4H j

[0106] Among them, Y represents the threat level of fire, β1, β2, β3, β4 are weight coefficients, T j is the measured temperature at point j, W fj is the wind speed measured at point j, P djis the precipitation measured at point j, H j is the relative humidity measured at point j. j , W fj , P dj , H j The average values ​​of the data measured by all sensors in the area are obtained and then normalized.

[0107] When Y is greater than the set threshold, it means that the fire threat level in the area is high and it is a potential fire point. If Y is greater than the set threshold and the above threshold conditions are met at the same time, it is judged as a potential fire point.

[0108] S2: Extract the absolute fire point from the potential fire points according to the brightness temperature of the third channel;

[0109] In non-cloud pixels, the brightness temperature of the third channel is 330K or above. If T3>330 during the day and T3>330 at night, the pixel is a fire point. T3 is the brightness temperature of the third channel, and the value of T3 is adjusted according to different regions and seasons.

[0110] The absolute fire point extraction method can also be performed according to the above method, combined with Y and threshold judgment conditions.

[0111] S3: Extract possible fire points other than absolute fire points from potential fire points according to the set discrimination conditions.

[0112] The specific method is as follows:

[0113] Non-cloud pixels in the forest fire concern area and non-absolute fire point pixels meet A or B and C:

[0114] A: T3-T 3b >4δT 3b

[0115] B: T3-T 3b >ΔT3

[0116] C: T3-T4-T 34b >4δT 34b

[0117] Among them, T3 is the brightness temperature value of the 3rd channel, T4 is the brightness temperature value of the 4th channel, T 3b is the average brightness temperature of the background pixels in channel 3, T 34b is the background temperature mean of the brightness temperature difference between channels 3 and 4, δT 3b is the standard deviation of the brightness temperature of the background pixels in channel 3, δT 34b is the background temperature standard deviation of the brightness temperature difference between the 3rd and 4th channels, and ΔT3 is the brightness temperature anomaly of the 3rd channel. 3b , δT 34bWhen the standard deviation is less than 2 K, set it equal to 2 K. ΔT3, when the daytime temperature is less than 15 K in the summer half year (May-October), take 15 K, and when the daytime temperature is less than 10 K in the winter half year (September-April of the following year), take 10 K. When the nighttime temperature is less than 10 K in the summer half year (May-October), take 10 K, and when the nighttime temperature is less than 8 K in the winter half year (September-April of the following year), take 8 K.

[0118] When calculating background information, it is necessary to ensure that the number of valid pixels is not less than 25% of the total number of pixels in the background area. If the adjacent area does not meet this condition, the range will be automatically expanded and the value will be re-taken. When the search range is expanded, the search is performed from near to far. The search pixel should be the same as the current discrimination pixel type. The maximum search range is 21×21 grid points. When the valid background pixels obtained when reaching 21×21 grid points still do not meet the conditions, the current detection pixel is considered to be a non-fire point. The background pixel is required to be a non-cloud pixel in the forest fire concern area.

[0119] A verification method for a spatiotemporal adaptive forest fire monitoring method based on FY3C data was proposed, which compared the identified potential fire points with the real fire points for verification, including the verification of forest fire location and forest fire area.

[0120] Methods for checking forest fire location Figure 1 As shown below:

[0121] If the monitored forest fire point falls within the real forest fire pixel or at a set position around the real forest fire pixel, the monitored forest fire point is consistent with the actual forest fire point.

[0122] X i,j 、F n,m are the locations of the real forest fire and the detected forest fire respectively; i and j are the latitude and longitude of the real forest fire respectively. n and m are the latitude and longitude of the detected forest fire respectively; when point X i,j 、F n,m When the map has the same coordinate system, the forest fire pixel F n,m Falls within the following pixels: X i-0.01,j-0.01 or X i-0.01,j or X i-0.01,j+0.01 or X i,j-0.01 or X i,j+0.01 or X i-0.01,j-0.01 or X i+0.01,j or X i+0.01,j+0.01 ; If it falls within the above-mentioned pixels, the detected forest fire location is correct.

[0123] The method for checking forest fire area is as follows:

[0124] Open fire area S Fde and the actual forest fire area S Fac If the area ratio is within the set threshold, the monitored forest fire area is considered to have passed the inspection;

[0125]

[0126] Where S Fac is the actual forest fire area. The pixel emissivity observed by the satellite is the weighted average of the emissivity of all parts of the ground within the coverage of the pixel. When a fire point appears on the ground, the pixel is divided into an open fire area and a background area (unburned area). The pixel containing the fire point is called a mixed pixel. It is assumed that the radiation characteristics of the open fire area, mixed pixel, and background area in the mid-infrared and far-infrared are black bodies.

[0127] S Fde =P×S

[0128] Among them, P is the percentage of the open fire area to the pixel area, and S is the area of ​​a single pixel;

[0129] P can be obtained as follows:

[0130]

[0131] Among them, N if is the mixed pixel radiance, N ihi is the emissivity of the open fire area, N ibg is the radiation rate of the background area around the fire point, T hi is the temperature of the open flame zone, C1 and C2 are constants, T bg is the background temperature; i is the channel number, V i is the central wave number of channel i. The first radiation constant C1=3.7427×10 8 W·μm 4 ·m -2 , the second radiation constant T bg The background temperature can be approximated by the non-fire point pixels around the mixed pixel. if Mixed pixel radiance can be obtained on remote sensing images.

[0132] P and T hi According to Newton's iteration method, the following equations are solved:

[0133]

[0134] The percentage P of the open fire area to the pixel area is obtained through the above equations, and the area S of a single pixel is obtained according to the pixel resolution.

[0135] Figure 3 The original images of the 3rd (left) and 4th (right) channels of FY3C are used to monitor forest fires using the method of the present invention and the traditional method. The results are as follows: Figure 4As shown in the figure, the left side shows the monitoring results obtained by the traditional method, and the right side shows the monitoring results obtained by the method of the present invention. It can be seen from the figure that the method of the present invention eliminates most high-temperature hot spots in non-cloud areas and improves the monitoring accuracy.

[0136] The present invention identifies hot spots through the threshold method, determines the threshold value according to the changes in region, time and underlying surface, and avoids the defects caused by too high or too low thresholds; uses surface cover data to solve the interference of other ground heat sources on forest fire monitoring, can effectively identify forest fire hot spots, and accurately distinguish fire points. The judgment result is more accurate when combined with the sensor test results, and the processing efficiency is also improved by processing data through neural networks.

Claims

1. A spatiotemporal adaptive forest fire monitoring method based on FY3C data, characterized in that: The following steps are involved: Step 1: Obtain the required data and preprocess the data; Step 2: Identify forest fire concern areas using the binary method based on the preprocessed data; the method for extracting forest fire concern areas is as follows: Identify non-forest fire concern areas and mark them with a value of 1, and the rest of the areas are forest fire concern areas and are marked with a value of 0; use the forest fire concern area and non-concern area binary layers as masks to extract the forest fire concern areas used to detect the image; Among them, the non-forest fire concern area extraction method extracts urban and rural construction land, road land, industrial and mining land, paddy fields, and fixed water bodies based on surface cover data; detects variable water bodies based on remote sensing images; variable water bodies meet the following conditions: Pixels meet and in, R 1 is the reflectivity of the first channel, R 2 is the reflectivity of the second channel, α 1 and α 2 is the threshold setting; Step 3: Identify cloud pixels from the forest fire concern area obtained in step 2; Step 4: Perform forest fire monitoring and identification on the non-cloud pixels in the forest fire focus area in step 2; first divide the forest fire monitoring units and monitoring sub-areas before forest fire identification; establish a brightness temperature query library based on the forest fire monitoring units and monitoring sub-areas, and then perform forest fire identification; The division method of forest fire monitoring units and monitoring sub-areas is as follows: The data is partitioned according to the latitude and longitude range and terrain. Each scanned image is divided into an independent monitoring unit according to the range of 1°×1°, and the monitoring sub-area is divided according to the elevation setting threshold within the monitoring unit; The method for establishing the brightness temperature query library is as follows: The 4th channel is used to calculate and build tables for the long-term multi-period monitoring surface brightness temperature values ​​of the forest fire monitoring unit and the monitoring sub-area; the long-term series duration is set according to needs, and the period length is set to one quarter or one month. The mean daytime brightness temperature and the mean nighttime brightness temperature are calculated in the period; the brightness temperature query database of different monitoring sub-areas in the monitoring unit is established based on the calculated data; The identification process is as follows: S1: Extract abnormally high temperature points from non-cloud pixels in the forest fire concern area as potential fire points; S2: Extract the absolute fire point from the potential fire points according to the brightness temperature of the third channel; S3: According to the set discrimination conditions, possible fire points other than the absolute fire points are extracted from the potential fire points; the specific method is as follows: Non-cloud pixels in the forest fire concern area and non-absolute fire point pixels meet A or B and C: A： B： C： in, T 3 is the brightness temperature value of the 3rd channel, T 4 is the brightness temperature value of the 4th channel, T 3b is the average brightness temperature of the background pixels in channel 3, T 34b is the background temperature mean of the brightness temperature difference between channels 3 and 4, is the standard deviation of the brightness temperature of the background pixels in channel 3, is the background temperature standard deviation of the brightness temperature difference between channels 3 and 4, is the brightness temperature anomaly of channel 3.

2. The spatiotemporal adaptive forest fire monitoring method based on FY3C data according to claim 1, characterized in that: The method for identifying cloud pixels in step 3 is as follows: If a pixel meets the following conditions during the day, it is a cloud pixel: and Or meet the following conditions: and If the pixel satisfies at night: in, R 1 is the reflectivity of the first channel, R 2 is the reflectivity of the second channel, T 5 is the brightness temperature value of the fifth channel, α 3. α 4. α 5. α 6. α 7 is the threshold setting.

3. The spatiotemporal adaptive forest fire monitoring method based on FY3C data according to claim 1, characterized in that: The method for extracting abnormally high temperature points in step S1 is as follows: The non-cloud pixels in the forest fire concern area meet the requirements, and the brightness temperature of the third channel is greater than the set threshold T h , the brightness temperature of the 4th channel is greater than the value of the brightness temperature query library for this period, and the reflectivity of the 1st and 2nd channels is less than the set threshold at the same time.

4. The spatiotemporal adaptive forest fire monitoring method based on FY3C data according to claim 1, characterized in that: The method of extracting the absolute fire point according to the brightness temperature of the third channel in step S2 is as follows: For non-cloud pixels in the forest fire concern area, the brightness temperature of the third channel must be greater than the set threshold during the day or the set threshold at night.

5. The testing method of any one of the spatiotemporal adaptive forest fire monitoring methods based on FY3C data as claimed in claims 1 to 4, characterized in that: The identified potential fire points are compared with the actual fire points for verification, including the inspection of forest fire location and forest fire area; the forest fire area inspection method is as follows: Open fire area S Fde and actual forest fire area S Fac If the area ratio is within the set threshold, the monitored forest fire area is considered to have passed the inspection; in, P is the percentage of open fire area to pixel area, S is the area of ​​a single pixel; in P It can be obtained as follows: in, N if is the mixed pixel radiance, N ihi is the emissivity of the open fire area, N ibg is the radiation rate of the background area around the fire point, T hi is the temperature of the open fire zone, C 1 and C 2 is a constant, T bg is the background temperature; i is the channel number, V i For Channel i The central wave number of P and T hi Solved by Newton's iteration method.

6. The verification method of a spatiotemporal adaptive forest fire monitoring method based on FY3C data according to claim 5, characterized in that: The forest fire location inspection method is as follows: If the monitored forest fire point falls within the real forest fire pixel or at a set position around the real forest fire pixel, the monitored forest fire point is consistent with the actual forest fire point.

Images