A fire monitoring method and terminal based on satellite remote sensing

Abstract

The present invention discloses a fire monitoring method based on satellite remote sensing, which includes acquiring satellite remote sensing data, performing pixel identification on the satellite remote sensing data, obtaining residual pixels and redundant pixels where fire is unlikely to exist; performing weighted convolution filtering on the satellite remote sensing data according to the redundant pixels and the residual pixels, obtaining filtered satellite remote sensing data; fitting a preset sample set using a linear regression model to obtain a preset brightness temperature value prediction model, and performing potential fire point identification based on the residual pixels in the filtered satellite remote sensing data and the preset brightness temperature value prediction model to obtain potential fire point pixels; determining target fire point pixels based on the potential fire point pixels, and determining a fire event based on the target fire point pixels, enabling each pixel to adapt to a time mode to have its own independent monitoring threshold, accurately identifying potential fire points, improving the accuracy of potential fire point identification, and thus improving the accuracy of fire monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of disaster prevention monitoring, and in particular to a fire monitoring method and terminal based on satellite remote sensing. Background Art

[0002] In recent years, forest fire disasters have occurred frequently. It is a natural disaster that is sudden, destructive, and difficult to handle and rescue. It has caused irreversible damage to the natural ecosystem and human life and property. Using satellite remote sensing to monitor forest fires is an effective technical means with a wide monitoring range, short monitoring cycle, and fewer restrictions.

[0003] Current satellite remote sensing methods for forest fire monitoring include noise removal and potential fire spot identification. The noise removal step typically smooths the entire image using methods such as mean filtering. This can blur the characteristics of fire spots near cloud and water points, leading to missed detections and reduced fire monitoring accuracy. The potential fire point identification step is to filter out pixels that are obviously not fire points based on the initial threshold. This step is the most basic and crucial step in forest fire identification. Due to the large difference in brightness temperature values ​​of satellite remote sensing data during the day and at night, the conventional forest fire monitoring method based on geosynchronous orbit satellites will set two fixed initial thresholds for day and night. However, there is not a hard transition between day and night but a smooth transition. In addition, my country has a vast territory and there is a large time difference between the east and west. If the initial threshold under the appropriate time mode is not selected, it is easy to cause false alarms and missed alarms, which further leads to a decrease in the accuracy of fire monitoring. In addition, the results of existing fire monitoring products are displayed in the form of fire points, but forest fires are characterized by a large area and long duration. A forest fire may have dozens or even hundreds of fire points. This display method is not convenient for users to view. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a fire monitoring method and terminal based on satellite remote sensing, which can improve the accuracy of fire monitoring.

[0005] In order to solve the above technical problems, a technical solution adopted by the present invention is:

[0006] A fire monitoring method based on satellite remote sensing, comprising:

[0007] Obtaining satellite remote sensing data and performing pixel recognition on the satellite remote sensing data to obtain remaining pixels and redundant pixels where fire is unlikely to exist;

[0008] performing weighted convolution filtering on the satellite remote sensing data according to the redundant pixels and the remaining pixels to obtain filtered satellite remote sensing data;

[0009] Using a linear regression model to fit a preset sample set to obtain a preset brightness temperature prediction model, and performing potential fire point identification based on the remaining pixels in the filtered satellite remote sensing data and the preset brightness temperature prediction model to obtain potential fire point pixels;

[0010] Target fire point pixels are determined based on the potential fire point pixels, and fire events are determined according to the target fire point pixels.

[0011] In order to solve the above technical problems, another technical solution adopted by the present invention is:

[0012] A fire monitoring terminal based on satellite remote sensing includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the following steps are implemented:

[0013] Obtaining satellite remote sensing data and performing pixel recognition on the satellite remote sensing data to obtain remaining pixels and redundant pixels where fire is unlikely to exist;

[0014] performing weighted convolution filtering on the satellite remote sensing data according to the redundant pixels and the remaining pixels to obtain filtered satellite remote sensing data;

[0015] Using a linear regression model to fit a preset sample set to obtain a preset brightness temperature prediction model, and performing potential fire point identification based on the remaining pixels in the filtered satellite remote sensing data and the preset brightness temperature prediction model to obtain potential fire point pixels;

[0016] Target fire point pixels are determined based on the potential fire point pixels, and fire events are determined according to the target fire point pixels.

[0017] The beneficial effects of the present invention are as follows: pixel identification is performed on the acquired satellite remote sensing data, the remaining pixels and redundant pixels where fire is unlikely to exist are identified, weighted convolution filtering is performed on the satellite remote sensing data according to the redundant pixels and the remaining pixels, potential fire points are identified based on the remaining pixels in the filtered satellite remote sensing data and a preset brightness temperature value prediction model, target fire point pixels are determined based on the identified potential fire point pixels, and fire events are determined based on the target fire point pixels. Unlike the prior art, noise removal is no longer performed by using mean filtering, but by using weighted convolution filtering after pixel identification to avoid the situation where the fire point characteristics in the noise adjacent area are not obvious after filtering. In addition, when performing potential fire point identification, the preset brightness temperature value prediction model used is fitted by a linear regression model, so that each pixel can adapt to the time mode to have its own independent monitoring threshold. Even at dawn and dusk, potential fire points can be accurately identified, thereby improving the accuracy of potential fire point identification and thus improving the accuracy of fire monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A flowchart of a fire monitoring method based on satellite remote sensing according to an embodiment of the present invention;

[0019] Figure 2 Schematic diagram of the structure of a fire monitoring terminal based on satellite remote sensing according to an embodiment of the present invention;

[0020] Figure 3 The figure is a flow chart of determining a fire event in a fire monitoring method based on satellite remote sensing according to an embodiment of the present invention. DETAILED DESCRIPTION

[0021] To illustrate the technical content, achieved objectives and effects of the present invention in detail, the following description is given in conjunction with the embodiments and accompanying drawings.

[0022] Please refer to Figure 1 The embodiment of the present invention provides a fire monitoring method based on satellite remote sensing, comprising:

[0023] Obtaining satellite remote sensing data and performing pixel recognition on the satellite remote sensing data to obtain remaining pixels and redundant pixels where fire is unlikely to exist;

[0024] performing weighted convolution filtering on the satellite remote sensing data according to the redundant pixels and the remaining pixels to obtain filtered satellite remote sensing data;

[0025] Using a linear regression model to fit a preset sample set to obtain a preset brightness temperature prediction model, and performing potential fire point identification based on the remaining pixels in the filtered satellite remote sensing data and the preset brightness temperature prediction model to obtain potential fire point pixels;

[0026] Target fire point pixels are determined based on the potential fire point pixels, and fire events are determined according to the target fire point pixels.

[0027] From the above description, it can be seen that the beneficial effects of the present invention are: pixel recognition is performed on the acquired satellite remote sensing data, the remaining pixels and redundant pixels where fire is unlikely to exist are identified, weighted convolution filtering is performed on the satellite remote sensing data based on the redundant pixels and the remaining pixels, potential fire points are identified based on the remaining pixels in the filtered satellite remote sensing data and a preset brightness temperature value prediction model, target fire point pixels are determined based on the identified potential fire point pixels, and fire events are determined based on the target fire point pixels. Unlike the prior art, mean filtering is no longer used for noise removal. Instead, weighted convolution filtering is used for noise removal after pixel recognition to avoid the situation where the fire point characteristics in the noise-adjacent area are not obvious after filtering. In addition, when identifying potential fire points, the preset brightness temperature value prediction model used is fitted by a linear regression model, which enables each pixel to adapt to the time mode to have its own independent monitoring threshold. Even at dawn and dusk, potential fire points can be accurately identified, thereby improving the accuracy of potential fire point identification and thus improving the accuracy of fire monitoring.

[0028] Furthermore, the redundant pixels include cloud point pixels and water point pixels;

[0029] The pixel identification of the satellite remote sensing data to obtain the remaining pixels and the redundant pixels where fire is unlikely to exist includes:

[0030] Obtaining, based on the satellite remote sensing data, a reflectivity of a second preset central wavelength, a reflectivity of a third preset central wavelength, a reflectivity of a fourth preset central wavelength, a brightness temperature value of a fifteenth preset central wavelength, and a solar altitude angle corresponding to each pixel;

[0031] Based on the reflectivity of the third preset central wavelength, the reflectivity of the fourth preset central wavelength, the brightness temperature value of the fifteenth preset central wavelength, and the solar altitude angle, a first recognition formula is used to perform cloud point pixel recognition to obtain cloud point pixels;

[0032] Based on the reflectivity of the second preset central wavelength, the reflectivity of the fourth preset central wavelength and the solar altitude angle, a second identification formula is used to identify water point pixels to obtain water point pixels;

[0033] The remaining pixels are determined according to the cloud point pixels and the water point pixels.

[0034] Furthermore, the first identification formula label_cloud(i) is:

[0035]

[0036] Where x 3i represents the reflectivity of the third preset central wavelength corresponding to the i-th pixel, x 4i represents the reflectivity of the fourth preset central wavelength corresponding to the i-th pixel, x15i represents the brightness temperature value of the fifteenth preset central wavelength corresponding to the i-th pixel, x θi represents the solar altitude angle corresponding to the i-th pixel;

[0037] The second identification formula label_water(i) is:

[0038]

[0039] Where x 2i represents the reflectivity of the second preset central wavelength corresponding to the i-th pixel.

[0040] From the above description, it can be seen that under natural conditions, it is impossible for a fire to exist in a location with water, and a fire cannot be effectively detected in a location covered by clouds, that is, corresponding to water point pixels and cloud point pixels, the remaining pixels are pixels where fire may exist. The first recognition formula is used to identify cloud point pixels based on the reflectivity of the third preset central wavelength, the reflectivity of the fourth preset central wavelength, the brightness temperature value of the fifteenth preset central wavelength, and the solar altitude angle. The second recognition formula is used to identify water point pixels based on the reflectivity of the second preset central wavelength, the reflectivity of the fourth preset central wavelength, and the solar altitude angle. This can accurately and effectively exclude cloud point pixels and water point pixels, reduce interference, and reduce the computational complexity of subsequent fire point detection algorithms.

[0041] Furthermore, performing weighted convolution filtering on the satellite remote sensing data according to the redundant pixels and the remaining pixels to obtain filtered satellite remote sensing data includes:

[0042] Obtaining a brightness temperature value of a seventh preset central wavelength corresponding to each pixel in the satellite remote sensing data;

[0043] Using a filter kernel to filter out the water point pixels and the cloud point pixels according to the brightness temperature value of the seventh preset central wavelength, to obtain filtered satellite remote sensing data;

[0044] Convolution filtering is performed on each remaining pixel in the filtered satellite remote sensing data according to the brightness temperature value of the seventh preset central wavelength to obtain filtered satellite remote sensing data.

[0045] From the above description, it can be seen that the filter kernel is used to filter out water point pixels and cloud point pixels in the satellite remote sensing data according to the brightness temperature value of the seventh preset central wavelength, and the water point pixels and cloud point pixels are excluded and not included in the subsequent calculations. Then, convolution filtering is performed on each remaining pixel in the filtered satellite remote sensing data according to the brightness temperature value of the seventh preset central wavelength, thereby realizing weighted convolution filtering, which can effectively remove the influence of cloud point pixels and water point pixels, and is conducive to improving the accuracy of fire monitoring.

[0046] Furthermore, the potential fire point identification is performed based on the remaining pixels in the filtered satellite remote sensing data and the preset brightness temperature value prediction model, and the potential fire point pixels obtained include:

[0047] Obtaining the solar altitude angle, the brightness temperature value of the fourteenth preset central wavelength, and the brightness temperature value of the seventh preset central wavelength corresponding to the remaining pixels according to the satellite remote sensing data;

[0048] Calculating using the preset brightness temperature prediction model according to the solar altitude angle corresponding to the remaining pixels to obtain predicted brightness temperature values ​​corresponding to the remaining pixels;

[0049] The third identification formula is used to identify potential fire points for the remaining pixels according to the predicted brightness temperature value, the brightness temperature value of the fourteenth preset central wavelength, and the brightness temperature value of the seventh preset central wavelength to obtain potential fire point pixels.

[0050] Furthermore, the third identification formula label_potential is:

[0051]

[0052] Where x 7i represents the brightness temperature value of the seventh preset central wavelength corresponding to the i-th remaining pixel, Y xθi represents the predicted brightness temperature value corresponding to the i-th remaining pixel, x 14i represents the brightness temperature value of the fourteenth preset central wavelength corresponding to the i-th remaining pixel.

[0053] From the above description, it can be seen that the preset brightness temperature prediction model is used to calculate according to the solar altitude angle corresponding to the remaining pixels to obtain the predicted brightness temperature values ​​corresponding to the remaining pixels, so that each remaining pixel can have its own independent detection threshold in the adaptive time mode even in the dawn and dusk periods. The third identification formula is used to identify potential fire points of the remaining pixels based on the predicted brightness temperature value, the brightness temperature value of the fourteenth preset central wavelength, and the brightness temperature value of the seventh preset central wavelength, thereby improving the accuracy of potential fire point identification and thus improving the accuracy of fire monitoring.

[0054] Furthermore, determining the target fire point pixel based on the potential fire point pixel includes:

[0055] Determining a background window corresponding to the potential fire point pixel;

[0056] Obtaining the remaining pixels in the background window except the potential fire point pixel, the brightness temperature values ​​of the seventh preset central wavelength corresponding to the remaining pixels, and the brightness temperature values ​​of the fourteenth preset central wavelength corresponding to the remaining pixels;

[0057] Calculating according to the brightness temperature values ​​of the seventh preset central wavelength corresponding to the remaining pixels, to obtain an average brightness temperature of the seventh preset central wavelength corresponding to the remaining pixels and a brightness temperature variance of the seventh preset central wavelength;

[0058] Calculating based on the brightness temperature value of the seventh preset central wavelength corresponding to the remaining pixel and the brightness temperature value of the fourteenth preset central wavelength corresponding to the remaining pixel to obtain a difference, an average value corresponding to the difference, and a variance corresponding to the difference;

[0059] Determine the initial target fire point pixel according to the average brightness temperature of the seventh preset central wavelength corresponding to the remaining pixels, the brightness temperature variance of the seventh preset central wavelength, the difference, the average corresponding to the difference, and the variance corresponding to the difference;

[0060] False fire points are eliminated from the initial target fire point pixels to obtain target fire point pixels.

[0061] From the above description, it can be seen that the remaining pixels are valid pixels, that is, pixels that are not clouds, water, or potential fire points. The initial target fire point pixels are determined based on the average brightness temperature of the seventh preset central wavelength corresponding to the remaining pixels, the brightness temperature variance of the seventh preset central wavelength, the difference, the average value corresponding to the difference, and the variance corresponding to the difference. The false fire points of the initial target fire point pixels are eliminated to obtain the target fire point pixels, which can achieve further screening of potential fire point pixels and ensure the validity of the target fire point pixels finally obtained.

[0062] Furthermore, the removing of false fire points from the initial target fire point pixels to obtain target fire point pixels comprises:

[0063] Obtaining the reflectivity of the first preset central wavelength, the reflectivity of the second preset central wavelength, the reflectivity of the third preset central wavelength, the reflectivity of the fourth preset central wavelength and the solar altitude angle corresponding to the initial target fire point pixel;

[0064] Determine whether the reflectivity of the first preset center wavelength is greater than the first preset value, whether the reflectivity of the second preset center wavelength is greater than the second preset value, whether the reflectivity of the third preset center wavelength is greater than the third preset value, whether the reflectivity of the fourth preset center wavelength is greater than the fourth preset value, and whether the solar altitude angle is less than the fifth preset value. If all of them are true, the initial target fire point pixel is determined as a false fire point; if not, the initial target fire point pixel is determined as a target fire point pixel.

[0065] From the above description, it can be seen that it is necessary to judge whether the reflectivity of the first preset central wavelength is greater than the first preset value, whether the reflectivity of the second preset central wavelength is greater than the second preset value, whether the reflectivity of the third preset central wavelength is greater than the third preset value, whether the reflectivity of the fourth preset central wavelength is greater than the fourth preset value, and whether the solar altitude angle is less than the fifth preset value. If all of them are true, the initial target fire point pixel is determined as a false fire point, which simply and effectively realizes the elimination of false fire points, thereby improving the accuracy of fire monitoring.

[0066] Furthermore, determining the fire event according to the target fire point pixel includes:

[0067] Obtaining the current moment corresponding to the target fire point pixel, and determining the target fire point pixel corresponding to the same current moment, to obtain the target fire point pixel to be merged;

[0068] Merging the target fire point pixels to be merged according to the eight-neighborhood connectivity method to obtain a first fire point set;

[0069] Determine the next moment corresponding to the current moment and the second fire point set corresponding to the next moment, and judge whether the intersection of the second fire point set and the first fire point set is empty. If not, merge the second fire point set with the first fire point set to obtain a fire event, and determine the next moment as the current moment, determine the second fire point set as the first fire point set, return to execute the step of determining the next moment corresponding to the current moment and the second fire point set corresponding to the next moment. If so, determine the first fire point set as a fire event.

[0070] From the above description, it can be seen that the detected fire points (i.e., target fire point pixels) are merged in the spatial dimension using the eight-neighborhood connectivity method and in the temporal dimension using the intersection detection method to form fire events. This avoids the problem of being inconvenient to view due to the large number of points and high repetition displayed in the form of fire points alone, making it easier to grasp the occurrence of fire events more intuitively.

[0071] Please refer to Figure 2 Another embodiment of the present invention provides a fire monitoring terminal based on satellite remote sensing, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, each step of the above-mentioned fire monitoring method based on satellite remote sensing is implemented.

[0072] The satellite remote sensing-based fire monitoring method and terminal of the present invention can be applied to various types of fire monitoring, such as forest fires and grassland fires, and are described below through specific implementation methods:

[0073] Example 1

[0074] Referring to the figure, a fire monitoring method based on satellite remote sensing in this embodiment includes:

[0075] S1. Obtain satellite remote sensing data and perform pixel recognition on the satellite remote sensing data to obtain remaining pixels and redundant pixels where fire is unlikely to exist, specifically including:

[0076] The redundant pixels include cloud point pixels and water point pixels; the satellite remote sensing data can be collected using a geostationary meteorological satellite. In this embodiment, the Sunflower-8 satellite is used to collect the satellite remote sensing data;

[0077] S11. Obtaining, based on the satellite remote sensing data, a reflectivity of a second preset central wavelength, a reflectivity of a third preset central wavelength, a reflectivity of a fourth preset central wavelength, a brightness temperature value of a fifteenth preset central wavelength, and a solar altitude angle corresponding to each pixel;

[0078] Among them, the second preset central wavelength is 0.510 μm central wavelength, corresponding to the satellite remote sensing data collected by the second channel of the Sunflower-8 satellite, the third preset central wavelength is 0.645 μm central wavelength, corresponding to the satellite remote sensing data collected by the third channel of the Sunflower-8 satellite, the fourth preset central wavelength is 0.860 μm central wavelength, corresponding to the satellite remote sensing data collected by the fourth channel of the Sunflower-8 satellite, and the fifteenth preset central wavelength is 12.350 μm central wavelength, corresponding to the satellite remote sensing data collected by the fifteenth channel of the Sunflower-8 satellite;

[0079] If other satellites are used for data collection, the corresponding channel is selected according to the different central wavelengths collected by different satellite channels;

[0080] S12, using a first recognition formula to identify cloud point pixels based on the reflectivity of the third preset central wavelength, the reflectivity of the fourth preset central wavelength, the brightness temperature value of the fifteenth preset central wavelength, and the solar altitude angle, to obtain cloud point pixels;

[0081] Among them, the first identification formula label_cloud(i) is:

[0082]

[0083] Where x 3i represents the reflectivity of the third preset central wavelength corresponding to the i-th pixel, x 4i represents the reflectivity of the fourth preset central wavelength corresponding to the i-th pixel, x 15i represents the brightness temperature value of the fifteenth preset central wavelength corresponding to the i-th pixel, x θi represents the solar altitude angle corresponding to the i-th pixel;

[0084] S13, identifying water point pixels using a second identification formula based on the reflectivity of the second preset central wavelength, the reflectivity of the fourth preset central wavelength, and the solar altitude angle to obtain water point pixels;

[0085] Among them, the second identification formula label_water(i) is:

[0086]

[0087] Where x 2i represents the reflectivity of the second preset central wavelength corresponding to the i-th pixel;

[0088] S14, determining the remaining pixels according to the cloud point pixels and the water point pixels;

[0089] Specifically, the pixels except cloud point pixels and water point pixels among all collected pixels are the remaining pixels;

[0090] S2. Performing weighted convolution filtering on the satellite remote sensing data according to the redundant pixels and the remaining pixels to obtain filtered satellite remote sensing data, specifically comprising:

[0091] S21, obtaining a brightness temperature value of a seventh preset central wavelength corresponding to each pixel in the satellite remote sensing data;

[0092] The seventh preset central wavelength is 3.850 μm, corresponding to the satellite remote sensing data collected by the seventh channel of the Sunflower-8 satellite;

[0093] S22, filtering out the water point pixels and the cloud point pixels using a filter kernel according to the brightness temperature value of the seventh preset central wavelength to obtain filtered satellite remote sensing data;

[0094] The filter kernel uses a 3×3 sliding window to perform weighted filtering;

[0095] Specifically, the water point pixels and the cloud point pixels are filtered out using a filter kernel according to the brightness temperature value of the seventh preset central wavelength, and the remaining pixels are averaged to obtain the weighted filtered satellite remote sensing data, that is,

[0096] S23, performing convolution filtering on each remaining pixel in the filtered satellite remote sensing data according to the brightness temperature value of the seventh preset central wavelength to obtain filtered satellite remote sensing data;

[0097] Specifically, performing convolution filtering on the brightness temperature value of the seventh preset central wavelength of each remaining pixel in the filtered satellite remote sensing data to obtain filtered satellite remote sensing data;

[0098] The convolution filtering is:

[0099]

[0100] Where, T 7i represents the brightness temperature value of the seventh preset central wavelength of the i-th remaining pixel after filtering, x 7i represents the brightness temperature value of the seventh preset central wavelength of the i-th remaining pixel, n represents the number of the filter kernels, and m represents the filter kernel index;

[0101] S3. Using a linear regression model to fit a preset sample set to obtain a preset brightness temperature prediction model, and performing potential fire point identification based on the remaining pixels in the filtered satellite remote sensing data and the preset brightness temperature prediction model to obtain potential fire point pixels, specifically including:

[0102] S31, using a linear regression model to fit a preset sample set to obtain a preset brightness temperature value prediction model;

[0103] Wherein, the preset sample set is a sample set of non-cloud, non-water, and non-fire points;

[0104] The preset brightness temperature prediction model is: t =aθ+b; where θ represents the solar altitude angle at time t, Y t represents the predicted brightness temperature at time t, a represents the first parameter, and b represents the second parameter;

[0105] S32. Obtaining, based on the satellite remote sensing data, the solar altitude angle, the brightness temperature value of the fourteenth preset central wavelength, and the brightness temperature value of the seventh preset central wavelength corresponding to the remaining pixels;

[0106] The fourteenth preset central wavelength is 11.2 μm, corresponding to the satellite remote sensing data collected by the fourteenth channel of the Sunflower-8 satellite;

[0107] S33, calculating using the preset brightness temperature prediction model according to the solar altitude angles corresponding to the remaining pixels to obtain predicted brightness temperature values ​​corresponding to the remaining pixels;

[0108] Specifically, the solar altitude angles corresponding to the remaining pixels are substituted into the preset brightness temperature prediction model to obtain the predicted brightness temperature values ​​corresponding to the remaining pixels;

[0109] S34, identifying potential fire points on the remaining pixels using a third identification formula according to the predicted brightness temperature value, the brightness temperature value of the fourteenth preset central wavelength, and the brightness temperature value of the seventh preset central wavelength, to obtain potential fire point pixels;

[0110] The third identification formula label_potential is:

[0111]

[0112] Where x 7i represents the brightness temperature value of the seventh preset central wavelength corresponding to the i-th remaining pixel, Y xθi represents the predicted brightness temperature value corresponding to the i-th remaining pixel, x 14i represents the brightness temperature value of the fourteenth preset central wavelength corresponding to the i-th remaining pixel;

[0113] S4, determining target fire point pixels based on the potential fire point pixels, and determining a fire event based on the target fire point pixels, specifically including:

[0114] S41, determining a background window corresponding to the potential fire point pixel, specifically comprising:

[0115] S411, determining an initial background window corresponding to the potential fire point pixel, wherein the window size of the initial background window is a preset size;

[0116] Wherein, the preset size is 5×5;

[0117] S412, calculating the number of the remaining pixels in the initial background window excluding the potential fire point pixels, and determining whether the number reaches a preset ratio of the initial background window, if so, executing S4121, if not, executing S4122;

[0118] Wherein, the preset ratio is 25%;

[0119] S4121, determining the initial background window as the background window corresponding to the potential fire point pixel;

[0120] S4122: Enlarging the window size of the initial background window to obtain an enlarged initial background window, and returning to S412 until the window size of the initial background window is enlarged to a preset maximum size;

[0121] For example, the window size of the initial background window is expanded in the order of 7×7, 9×9, ..., 21×21;

[0122] Wherein, when the window size of the initial background window is expanded to a preset maximum size and the number still does not reach a preset proportion of the initial background window, the potential fire point pixel is discarded;

[0123] S42, obtaining the remaining pixels in the background window except the potential fire point pixel, the brightness temperature values ​​of the seventh preset central wavelength corresponding to the remaining pixels, and the brightness temperature values ​​of the fourteenth preset central wavelength corresponding to the remaining pixels;

[0124] S43: Calculate according to the brightness temperature values ​​of the seventh preset central wavelength corresponding to the remaining pixels to obtain an average brightness temperature value of the seventh preset central wavelength and a brightness temperature variance of the seventh preset central wavelength corresponding to the remaining pixels;

[0125] S44, calculating based on the brightness temperature values ​​of the seventh preset central wavelength corresponding to the remaining pixels and the brightness temperature values ​​of the fourteenth preset central wavelength corresponding to the remaining pixels to obtain a difference, an average value corresponding to the difference, and a variance corresponding to the difference;

[0126] S45, determining an initial target fire point pixel according to the average brightness temperature of the seventh preset central wavelength corresponding to the remaining pixels, the brightness temperature variance of the seventh preset central wavelength, the difference, the average corresponding to the difference, and the variance corresponding to the difference;

[0127] Specifically, it is determined whether the brightness temperature value of the seventh preset central wavelength corresponding to the remaining pixels, the average brightness temperature of the seventh preset central wavelength, the brightness temperature variance of the seventh preset central wavelength, the difference, the average corresponding to the difference, and the variance corresponding to the difference all meet the preset conditions; if so, the remaining pixels are determined as the initial target fire point pixels; if not, the remaining pixels are not the initial target fire point pixels;

[0128] The preset conditions are:

[0129]

[0130]

[0131] Where, represents the average brightness temperature of the seventh preset central wavelength, δx 7bg represents the brightness temperature variance of the seventh preset central wavelength, Δx 7_14i represents the difference, represents the average value corresponding to the difference, δΔx 7_14bg represents the variance corresponding to the difference;

[0132] S46, removing false fire points from the initial target fire point pixels to obtain target fire point pixels, specifically comprising:

[0133] S461, obtaining the reflectivity of the first preset central wavelength, the reflectivity of the second preset central wavelength, the reflectivity of the third preset central wavelength, the reflectivity of the fourth preset central wavelength and the solar altitude angle corresponding to the initial target fire point pixel;

[0134] The first preset central wavelength is 0.455 μm, corresponding to the satellite remote sensing data collected by the first channel of the Sunflower-8 satellite;

[0135] S462, determining whether the reflectivity of the first preset central wavelength is greater than a first preset value, whether the reflectivity of the second preset central wavelength is greater than a second preset value, whether the reflectivity of the third preset central wavelength is greater than a third preset value, whether the reflectivity of the fourth preset central wavelength is greater than a fourth preset value, and whether the solar altitude angle is less than a fifth preset value; if all are yes, executing S4621; if not, executing S4622;

[0136] Among them, the first preset value is 0.35, the second preset value is 0.35, the third preset value is 0.35, the fourth preset value is 0.4, and the fifth preset value is 30;

[0137] The reflectivity of the first preset central wavelength is greater than a first preset value, the reflectivity of the second preset central wavelength is greater than a second preset value, the reflectivity of the third preset central wavelength is greater than a third preset value, the reflectivity of the fourth preset central wavelength is greater than a fourth preset value, and the solar altitude angle is less than a fifth preset value, that is, x 1i >0.35&x 2i >0.35&x 3i >0.35&x 4i >0.4&x θi <30;

[0138] S4621, determining the initial target fire point pixel as a false fire point;

[0139] S4622, determining the initial target fire point pixel as the target fire point pixel;

[0140] S47, obtaining the current moment corresponding to the target fire point pixel, and determining the target fire point pixel corresponding to the same current moment, to obtain the target fire point pixel to be merged;

[0141] S48, merging the target fire point pixels to be merged according to the eight-neighborhood connectivity method to obtain a first fire point set;

[0142] For example, for the merged target fire pixel i, its 8-neighborhood N8(i) t If there are other target fire point pixels to be merged within the range of , they will be merged into the set A of target fire point pixels i to be merged. i Then, the target fire point pixel to be merged in the 8-neighborhood continues to search for other target fire point pixels to be merged in its 8-neighborhood, and this cycle continues until all connected pixels are merged, forming the first fire point set A related to the target fire point pixel i to be merged at time t it ;

[0143] S49, determining a subsequent moment corresponding to the current moment and a second fire point set corresponding to the subsequent moment, and determining whether the intersection of the second fire point set and the first fire point set is empty; if not, executing S491; if yes, executing S492;

[0144] For example, determine the A corresponding to time t+1 it+1 ,like Then execute S491;

[0145] S491: Merge the second fire point set with the first fire point set to obtain a fire event, determine the later moment as the current moment, determine the second fire point set as the first fire point set, and return to execute step S49;

[0146] When there is no target fire point pixel to be merged that intersects with the current fire point set at time t+k, the fire event is merged;

[0147] S492: Determine the first fire point set as a fire event;

[0148] In an optional embodiment, the step S4 includes:

[0149] S5. Determine the time corresponding to the earliest target fire point pixel appearing in the fire event as the fire start time, and determine the time corresponding to the latest target fire point pixel appearing as the fire end time;

[0150] S6. Obtain a true color image corresponding to the fire start time to the fire end time, merge the true color image, the fire start time, and the fire end time to obtain a fire event sequence diagram, and display the fire event sequence diagram;

[0151] By generating fire start time, fire end time, fire event time sequence diagram, etc., users can intuitively understand the occurrence and development of fire.

[0152] Example 2

[0153] Please refer to Figure 2 In this embodiment, a fire monitoring terminal based on satellite remote sensing includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, each step of the fire monitoring method based on satellite remote sensing in the first embodiment is implemented.

[0154] In summary, the present invention provides a fire monitoring method and terminal based on satellite remote sensing, which obtains satellite remote sensing data, and performs pixel recognition on the satellite remote sensing data to obtain remaining pixels and redundant pixels where fire is unlikely to exist; performs weighted convolution filtering on the satellite remote sensing data according to the redundant pixels and the remaining pixels to obtain filtered satellite remote sensing data; uses a linear regression model to fit a preset sample set to obtain a preset brightness temperature value prediction model, and performs potential fire point recognition based on the remaining pixels in the filtered satellite remote sensing data and the preset brightness temperature value prediction model to obtain potential fire point pixels; determines target fire point pixels based on the potential fire point pixels, and determines a fire event based on the target fire point pixels, and performs a fire event detection on the detected fire point (i.e., the target fire point pixel) in the In the spatial dimension, the eight-neighborhood connectivity method is used for merging, and in the temporal dimension, the intersection detection method is used for merging to form fire events. Finally, by generating the fire start time, fire end time, and fire event time series diagram, it is convenient to intuitively grasp the occurrence and development rules of the fire; after pixel identification, weighted convolution filtering is used to remove noise to avoid the situation where the fire point characteristics in the noise-adjacent area are not obvious after filtering; and when identifying potential fire points, the preset brightness temperature value prediction model is fitted by the linear regression model, which enables each pixel to adapt to the time pattern to have its own independent monitoring threshold. Even at dawn and dusk, potential fire points can be accurately identified, which improves the accuracy of potential fire point identification and thus improves the accuracy of fire monitoring.

[0155] The above descriptions are merely embodiments of the present invention and are not intended to limit the scope of the present invention. Any equivalent transformations made using the contents of the present invention's description and drawings, or directly or indirectly applied in related technical fields, are also included in the scope of the present invention's patent protection.

Claims

1. A fire monitoring method based on satellite remote sensing, characterized in that: include: Obtaining satellite remote sensing data and performing pixel recognition on the satellite remote sensing data to obtain remaining pixels and redundant pixels where fire is unlikely to exist; performing weighted convolution filtering on the satellite remote sensing data according to the redundant pixels and the remaining pixels to obtain filtered satellite remote sensing data; Using a linear regression model to fit a preset sample set to obtain a preset brightness temperature prediction model, and performing potential fire point identification based on the remaining pixels in the filtered satellite remote sensing data and the preset brightness temperature prediction model to obtain potential fire point pixels; Determining target fire point pixels based on the potential fire point pixels, and determining a fire event based on the target fire point pixels; The performing weighted convolution filtering on the satellite remote sensing data according to the redundant pixels and the remaining pixels to obtain filtered satellite remote sensing data includes: Obtaining a brightness temperature value of a seventh preset central wavelength corresponding to each pixel in the satellite remote sensing data; Using a filter kernel to filter out water point pixels and cloud point pixels according to the brightness temperature value of the seventh preset central wavelength, to obtain filtered satellite remote sensing data; performing convolution filtering on each remaining pixel in the filtered satellite remote sensing data according to the brightness temperature value of the seventh preset central wavelength to obtain filtered satellite remote sensing data; The potential fire point identification is performed based on the remaining pixels in the filtered satellite remote sensing data and the preset brightness temperature value prediction model, and the potential fire point pixels are obtained, including: Obtaining the solar altitude angle, the brightness temperature value of the fourteenth preset central wavelength, and the brightness temperature value of the seventh preset central wavelength corresponding to the remaining pixels according to the satellite remote sensing data; Calculating using the preset brightness temperature prediction model according to the solar altitude angle corresponding to the remaining pixels to obtain predicted brightness temperature values ​​corresponding to the remaining pixels; The third identification formula is used to identify potential fire points for the remaining pixels according to the predicted brightness temperature value, the brightness temperature value of the fourteenth preset central wavelength, and the brightness temperature value of the seventh preset central wavelength to obtain potential fire point pixels.

2. The fire monitoring method based on satellite remote sensing according to claim 1, characterized in that: The redundant pixels include cloud point pixels and water point pixels; The pixel identification of the satellite remote sensing data to obtain the remaining pixels and the redundant pixels where fire is unlikely to exist includes: Obtaining, based on the satellite remote sensing data, a reflectivity of a second preset central wavelength, a reflectivity of a third preset central wavelength, a reflectivity of a fourth preset central wavelength, a brightness temperature value of a fifteenth preset central wavelength, and a solar altitude angle corresponding to each pixel; Based on the reflectivity of the third preset central wavelength, the reflectivity of the fourth preset central wavelength, the brightness temperature value of the fifteenth preset central wavelength, and the solar altitude angle, a first recognition formula is used to perform cloud point pixel recognition to obtain cloud point pixels; Based on the reflectivity of the second preset central wavelength, the reflectivity of the fourth preset central wavelength and the solar altitude angle, a second identification formula is used to identify water point pixels to obtain water point pixels; The remaining pixels are determined according to the cloud point pixels and the water point pixels.

3. A fire monitoring method based on satellite remote sensing according to claim 2, characterized in that: The first identification formula label_cloud(i) is: ； Where x 3i represents the reflectivity of the third preset central wavelength corresponding to the i-th pixel, x 4i represents the reflectivity of the fourth preset central wavelength corresponding to the i-th pixel, x 15i represents the brightness temperature value of the fifteenth preset central wavelength corresponding to the i-th pixel, x θi represents the solar altitude angle corresponding to the i-th pixel; The second identification formula label_water(i) is: ； Where x 2i represents the reflectivity of the second preset central wavelength corresponding to the i-th pixel.

4. The fire monitoring method based on satellite remote sensing according to claim 1, characterized in that: The third identification formula label_potential is: ； Where x 7i represents the brightness temperature value of the seventh preset central wavelength corresponding to the i-th remaining pixel, Y xθi represents the predicted brightness temperature value corresponding to the i-th remaining pixel, x 14i represents the brightness temperature value of the fourteenth preset central wavelength corresponding to the i-th remaining pixel.

5. The fire monitoring method based on satellite remote sensing according to claim 1, characterized in that: Determining the target fire point pixel based on the potential fire point pixel includes: Determining a background window corresponding to the potential fire point pixel; Obtaining the remaining pixels in the background window except the potential fire point pixel, the brightness temperature values ​​of the seventh preset central wavelength corresponding to the remaining pixels, and the brightness temperature values ​​of the fourteenth preset central wavelength corresponding to the remaining pixels; Calculating according to the brightness temperature values ​​of the seventh preset central wavelength corresponding to the remaining pixels, to obtain an average brightness temperature of the seventh preset central wavelength corresponding to the remaining pixels and a brightness temperature variance of the seventh preset central wavelength; Calculating based on the brightness temperature value of the seventh preset central wavelength corresponding to the remaining pixel and the brightness temperature value of the fourteenth preset central wavelength corresponding to the remaining pixel to obtain a difference, an average value corresponding to the difference, and a variance corresponding to the difference; Determine the initial target fire point pixel according to the average brightness temperature of the seventh preset central wavelength corresponding to the remaining pixels, the brightness temperature variance of the seventh preset central wavelength, the difference, the average corresponding to the difference, and the variance corresponding to the difference; False fire points are eliminated from the initial target fire point pixels to obtain target fire point pixels.

6. A fire monitoring method based on satellite remote sensing according to claim 5, characterized in that: The removing of false fire points from the initial target fire point pixels to obtain target fire point pixels comprises: Obtaining the reflectivity of the first preset central wavelength, the reflectivity of the second preset central wavelength, the reflectivity of the third preset central wavelength, the reflectivity of the fourth preset central wavelength and the solar altitude angle corresponding to the initial target fire point pixel; Determine whether the reflectivity of the first preset center wavelength is greater than the first preset value, whether the reflectivity of the second preset center wavelength is greater than the second preset value, whether the reflectivity of the third preset center wavelength is greater than the third preset value, whether the reflectivity of the fourth preset center wavelength is greater than the fourth preset value, and whether the solar altitude angle is less than the fifth preset value. If all of them are true, the initial target fire point pixel is determined as a false fire point; if not, the initial target fire point pixel is determined as a target fire point pixel.

7. The fire monitoring method based on satellite remote sensing according to claim 1, characterized in that: Determining a fire event according to the target fire point pixel includes: Obtaining the current moment corresponding to the target fire point pixel, and determining the target fire point pixel corresponding to the same current moment, to obtain the target fire point pixel to be merged; Merging the target fire point pixels to be merged according to the eight-neighborhood connectivity method to obtain a first fire point set; Determine the next moment corresponding to the current moment and the second fire point set corresponding to the next moment, and judge whether the intersection of the second fire point set and the first fire point set is empty. If not, merge the second fire point set with the first fire point set to obtain a fire event, and determine the next moment as the current moment, determine the second fire point set as the first fire point set, return to execute the step of determining the next moment corresponding to the current moment and the second fire point set corresponding to the next moment. If so, determine the first fire point set as a fire event.

8. A fire monitoring terminal based on satellite remote sensing, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, each step of the fire monitoring method based on satellite remote sensing according to any one of claims 1 to 7 is implemented.

Images