Fire point adaptive method based on multi-source synchronous satellite data

Abstract

The invention discloses a fire point self-adaption method based on multi-source synchronous satellite data, which relates to the technical field of remote sensing and geographic information, and comprises the following steps of: aiming at satellite remote sensing images with different resolutions, respectively selecting fire point detection algorithms matched with the resolution characteristics of the satellite remote sensing images to independently finish preliminary detection of fire points, extracting fire point space distribution and feature data corresponding to respective resolution scale; preprocessing the extracted multi-source fire point spatial distribution and feature data, and establishing a multi-source fire point data set; according to the method, the scale consistency and fusion precision of multi-source satellite fire point detection are improved, and the problems of fire point boundary breakage, false splitting and the like caused by resolution difference are solved. Through scale difference quantification and machine learning evaluation, in combination with scale correction of high-resolution images, adaptive dynamic adjustment of fire point boundaries is realized, and the reliability of fire monitoring and emergency decision making is improved.

Description

Technical Field

[0001] The present invention relates to the field of remote sensing and geographic information technology, and particularly to a fire point adaptive method based on multi-source synchronous satellite data. Background Art

[0002] Fire point adaptation based on multi-source synchronous satellite data refers to using multi-source remote sensing data (such as infrared, visible light, thermal infrared, radar, etc.) obtained from multiple satellites of different types or orbits, and through synchronous correction and fusion of their time, space, spectral and other characteristics, dynamically and intelligently adjusting the judgment criteria for fire point extraction (such as thresholds, feature combinations, fire point discrimination models), so as to adaptively detect and identify surface fire points according to the characteristics of different scenarios and different data sources. This method can overcome the problems restricted by factors such as weather, observation angle, and time resolution of single satellite data, and improve the accuracy and robustness of fire point detection through multi-source complementarity, information enhancement and adaptive adjustment, especially suitable for fire monitoring tasks under complex terrains, seasonal changes or variable meteorology.

[0003] The prior art has the following deficiencies: In the prior art, in the process of fire point detection and identification based on multi-source synchronous satellite data, although some methods have tried to consider the scale differences between the spatial resolutions of different satellite images and adopted conventional means such as resolution normalization and image resampling for processing, there is still a general lack of dynamic adaptive adjustment ability for scale differences in the fusion of multi-source fire point observation information. Specifically, the multiple satellites participating in fire point monitoring usually include high-resolution, medium-resolution and low-resolution remote sensing satellites, and the span of their imaging spatial resolutions is relatively large. The resolution of high-resolution satellites can reach the 30-meter level, while the resolutions of medium- and low-resolution satellites are often 250 meters or even more than 1 kilometer. Since the spatial scale of fire points is usually small, and there are significant differences in the spatial scale, spectral characteristics and morphological features of fire points at different time phases and different resolutions, if there is no adaptive scale control based on the observation scenario, fire point characteristics and image fusion process, the following two prominent problems are still likely to occur: On the one hand, the fire point areas detected in low-resolution remote sensing images are prone to be over-refined due to scale mapping distortion in high-resolution images, resulting in abnormal shrinkage, fragmentation or splitting of fire point boundaries, seriously affecting the accurate determination of the actual fire range, fire spread direction and situation; on the other hand, some weak fire points and edge fire points that are difficult to distinguish or merged in low-resolution images may be abnormally enlarged or misjudged as multiple independent strong fire points in high-resolution images, thus causing serious distortion of the fire point quantity and spatial distribution characteristics, and reducing the spatial consistency and authenticity of fire point detection and fire monitoring results.

[0004] The above information disclosed in the background art section is only used to enhance the understanding of the background of the present disclosure. Therefore, it may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0005] The object of the present invention is to provide a fire point adaptive method based on multi-source synchronous satellite data, which improves the scale consistency and fusion accuracy of multi-source satellite fire point detection, and solves problems such as fire point boundary fragmentation and false splitting caused by resolution differences. Through scale difference quantification and machine learning evaluation, combined with scale correction of high-resolution images, the adaptive dynamic adjustment of the fire point boundary is realized, and the reliability of fire monitoring and emergency decision-making is improved to solve the problems in the above background art.

[0006] To achieve the above object, the present invention provides the following technical solution: A fire point adaptive method based on multi-source synchronous satellite data, including the following steps: For satellite remote sensing images with different resolutions, fire point detection algorithms matching their resolution characteristics are respectively selected to independently complete the preliminary fire point detection, and the fire point spatial distribution and characteristic data corresponding to their respective resolution scales are extracted; Preprocess the extracted multi-source fire point spatial distribution and characteristic data, and establish a multi-source fire point data set; For the multi-source fire point detection results in the overlapping area, extract the scale difference information from the multi-source fire point data set, and quantitatively analyze the extracted scale difference information to comprehensively describe the observed difference characteristics of the fire point at different resolutions; Use the quantified scale difference information as a characteristic parameter, input it into a pre-trained machine learning model, and evaluate the scale error risk of the current multi-source fire point data through the machine learning model; When the model identifies the existence of scale error risk, taking the spatial scale and boundary characteristics of the high-resolution satellite data as a benchmark, adopt a scale consistency correction algorithm to perform scale mapping and boundary reconstruction on the low-resolution satellite fire point detection results, and dynamically correct the fire point area boundary and spatial distribution characteristics during the multi-source fire point fusion process to avoid the phenomenon of over-exaggeration or refinement of the fire point area after fusion.

[0007] Preferably, preprocessing the extracted multi-source fire point spatial distribution and characteristic data, and establishing a multi-source fire point data set specifically includes the following steps: First, perform time synchronization processing on the fire point spatial distribution and characteristic data from different satellites, and solve the timing error caused by inconsistent multi-source satellite observation times through observation time correction and interpolation completion; Secondly, perform unified spatial projection processing, and uniformly convert each source of data into the same geographic coordinate system and spatial resolution grid to ensure the corresponding relationship of the positions of each fire point in space; Again, perform noise removal, outlier detection and correction on the fire point feature data, and remove false fire points or distorted features caused by sensor errors and cloud and smoke interference; Then, for high-resolution, medium-resolution, and low-resolution satellite data, perform hierarchical storage and organization according to data sources, resolutions, and feature types, and establish a structured multi-source fire point data set; Finally, provide a standardized, spatio-temporally aligned, and feature-complete fire point observation data basis for subsequent scale difference analysis and fusion processing.

[0008] Preferably, for the multi-source fire point detection results in the overlapping area, extract scale difference information from the multi-source fire point data set. The extracted information includes the number of small fire points into which a single low-resolution fire point splits at high resolution and the difference in pixel area variance within the fire point area at different resolutions. Quantitatively analyze the number of small fire points into which a single low-resolution fire point splits at high resolution and the difference in pixel area variance within the fire point area at different resolutions, and generate a reference value for the fire point split number and a reference value for the local area variance respectively. Comprehensively describe the observed difference characteristics of the fire point at different resolutions through the reference value for the fire point split number and the reference value for the local area variance.

[0009] Preferably, the specific steps for quantitatively analyzing the number of small fire points into which a single low-resolution fire point splits at high resolution to generate a reference value for the fire point split number are as follows: Perform spatial matching on the low-resolution fire point area and the fire point set in the high-resolution image, and judge the spatial coverage and inclusion relationship. For each low-resolution fire point, search and identify all the included high-resolution fire points within the corresponding spatial range to obtain the corresponding number of high-resolution split fire points. The calculation expression is as follows: ; In the formula, is the fire point split number, indicating the number of effective high-resolution fire points into which the low-resolution fire point is split at high resolution, is the th high-resolution fire point, is the low-resolution fire point area, is the number of fire points in the high-resolution fire point candidate set, is the spatial inclusion indicator function; In order to characterize the severity of the splitting situation, combine the scale characteristics of the low resolution itself to generate a reference value for the fire point split number, measure the relationship between the number of split fire points and the low-resolution spatial scale, and the generation formula is as follows: ; In the formula, is the reference value for the fire point split number, is the boundary length of the low-resolution fire point, is the area of the low-resolution fire point.

[0010] Preferably, the specific steps for quantitatively analyzing the difference in pixel area variance within the extracted fire point region at different resolutions to generate a local area variance reference value are as follows: For the key fire point regions of the high-resolution image and the low-resolution image, respectively extract the set of pixel areas within the regions, denoted as and , is the set of pixel areas under the high-resolution image, is the set of pixel areas under the low-resolution image. Calculate the area distribution deviation value, and the calculation expression is as follows: ; In the formula, is the area distribution deviation value, and are respectively the maximum and minimum values of the pixel areas in the fire point region at high resolution, and are respectively the maximum and minimum values of the pixel areas in the fire point region at low resolution, is a very small positive value to prevent the denominator from being zero; Calculate the fractional spacing deviation of the pixel areas within the fire point region, and construct a local area variance reference value. The calculation expression is as follows: ; In the formula, is the local area variance reference value, and are respectively the 75th and 25th percentiles of the pixel areas at high resolution, and are respectively the 75th and 25th percentiles of the pixel areas at low resolution.

[0011] Preferably, take the quantified fire point split number reference value and the local area variance reference value as feature parameters, input them into a pre-trained machine learning model, generate a scale difference coefficient through the machine learning model, and evaluate the scale error risk of the current multi-source fire point data through the scale difference coefficient.

[0012] Preferably, compare and analyze the scale difference coefficient generated when evaluating the scale error risk of the current multi-source fire point data through a pre-trained machine learning model with a pre-set scale difference coefficient reference threshold to evaluate whether there is a scale error risk in the current multi-source fire point data. The specific steps are as follows: If the scale difference coefficient is greater than the pre-set scale difference coefficient reference threshold, it is determined that there is a scale error risk in the current multi-source fire point data; If the scale difference coefficient is less than or equal to a pre-set reference threshold of the scale difference coefficient, it is determined that there is no risk of scale error in the current multi-source hotspot data.

[0013] Preferably, when the model identifies a risk of scale error, based on the spatial scale and boundary characteristics of high-resolution satellite data, scale mapping and boundary reconstruction are performed on the low-resolution satellite hotspot detection results, and the specific steps for dynamically correcting the hotspot area boundary and spatial distribution characteristics are as follows: In the multi-source hotspot data set, taking the hotspot area of the low-resolution satellite as the reference area, the curvature distribution of the hotspot boundary and the area scale deformation factor within the area are extracted based on the high-resolution image to construct a scale difference field, and the formula is as follows: ; In the formula, is the difference intensity of the scale difference field at position , is the local curvature of the hotspot boundary in the high-resolution image at position , is the local curvature of the hotspot boundary in the low-resolution image at the corresponding position, is the local area scale function of the hotspot in the high-resolution image, is the local area scale function of the hotspot in the low-resolution image, is the curvature balance factor, is the area scale difference weight factor; Based on the difference intensity of the scale difference field, a scale consistency vector field is constructed to guide the scale consistency mapping of the low-resolution hotspot boundary points to the high-resolution reference boundary, and the formula is as follows: ; In the formula, is the scale consistency mapping vector field, is the gradient of the difference intensity of the scale difference field, is the normal vector of the high-resolution hotspot boundary, is the difference field gradient weight factor, is the high-resolution boundary normal component weight factor; Using the scale consistency mapping vector field to adjust the positions of the low-resolution hotspot boundary points, complete the boundary reconstruction, and complete the fusion of multi-source hotspots under the high-resolution reference. The calculation expression is as follows: ; In the formula, is the position of the corrected boundary point, is the original position of the low-resolution hotspot boundary point, is the step adjustment factor.

[0014] In the above technical solution, the technical effects and advantages provided by the present invention are as follows: The present invention effectively improves the spatial scale consistency and fusion accuracy of multi-source satellite fire point detection results, significantly improves problems such as fire point boundary fragmentation, false splitting, and abnormal area scaling caused by resolution differences, and ensures the accuracy and stability of the fused fire points in terms of spatial distribution, morphological characteristics, and scale information. Specifically, this solution dynamically extracts and quantifies scale difference information, combines a machine learning model to intelligently evaluate the scale error risk, and implements scale consistency correction based on high-resolution data as a benchmark, realizing the adaptive dynamic adjustment of multi-source fire point boundaries, effectively solving the problem of the lack of scale adaptability in multi-source fire point data fusion in the prior art, and improving the reliability and practicality of fire point detection results in fire monitoring, spread prediction, and emergency decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings in the following description are only some embodiments recorded in the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings.

[0016] Figure 1 is the method flow chart of the fire point self-adaptive method based on multi-source synchronous satellite data of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] Now, the exemplary embodiments will be described more comprehensively with reference to the drawings. However, the exemplary embodiments can be implemented in various forms and should not be construed as limited to the examples set forth herein; on the contrary, these exemplary embodiments are provided so that the present disclosure will be more thorough and complete, and the concept of the exemplary embodiments will be fully conveyed to those skilled in the art.

[0018] The present invention provides a Figure 1 fire point self-adaptive method based on multi-source synchronous satellite data as shown, including the following steps: For satellite remote sensing images with different resolutions, fire point detection algorithms matching their resolution characteristics are respectively selected to independently complete the preliminary fire point detection, and the fire point spatial distribution and characteristic data corresponding to their respective resolution scales are extracted; High-resolution satellites (such as 30m) are more suitable for the edge- and morphology-based hotspot detection method, which can refine the boundaries of small-scale hotspots; medium- and low-resolution satellites (such as 250m, 1km) are more suitable for using the detection method based on thermal anomalies or brightness temperature thresholds to identify large-scale high-temperature regions. The purpose of this step is to avoid the problem of insufficient adaptability caused by directly using a unified hotspot detection method to process data of different resolutions, and to give full play to the best capabilities of each resolution data for hotspot detection, laying a reliable single-source preliminary hotspot detection result for subsequent fusion.

[0019] After the preliminary hotspot detection is completed, extract the spatial distribution and characteristic data of hotspots from the detection results of each resolution, including but not limited to the location, boundary, area, morphology, brightness temperature, spectral characteristics, etc. of the hotspots. The purpose of this step is to establish the characteristics of the original observation data of multi-source hotspots and completely retain the characteristic differences of hotspots at different resolutions. Since the spatial and spectral characteristics of hotspots are different at high and low resolutions, only by completely extracting the characteristic data can it provide the necessary information support for subsequent scale difference analysis, risk assessment, and boundary reconstruction.

[0020] Preprocess the extracted multi-source hotspot spatial distribution and characteristic data, and establish a multi-source hotspot data set; Preprocess the extracted multi-source hotspot spatial distribution and characteristic data, and establish a multi-source hotspot data set, which specifically includes the following steps: First, perform time synchronization processing on the hotspot spatial distribution and characteristic data from different satellites, and solve the timing error caused by inconsistent observation times of multi-source satellites through observation time correction and interpolation; Second, perform unified spatial projection processing, and uniformly convert each source of data into the same geographic coordinate system and spatial resolution grid to ensure the corresponding relationship of each hotspot position in space; Third, perform noise removal, outlier detection and correction on the hotspot characteristic data (such as hotspot location, area, boundary shape, brightness temperature, spectral characteristics, etc.), and remove false hotspots or distorted characteristics caused by sensor errors and cloud and smoke interference; Fourth, for high-resolution, medium-resolution, and low-resolution satellite data, perform hierarchical storage and organization according to the data source, resolution, and characteristic type to establish a structured multi-source hotspot data set; Finally, provide a standardized, spatio-temporally aligned, and feature-complete hotspot observation data basis for subsequent scale difference analysis and fusion processing. Through the above preprocessing, the spatial consistency, time consistency, and feature reliability of multi-source hotspot data can be effectively improved, laying a foundation for achieving high-precision hotspot fusion and scale adaptability.

[0021] Preprocessing includes time synchronization, projection consistency correction, noise removal, data format standardization, etc., and uniformly constructs a multi-source fire point data set. The role of this step is to solve the inconsistencies in time, spatial reference system, and data format among different satellite images, ensuring that subsequent scale difference analysis and fire point fusion can be carried out under a unified framework. Preprocessing can also eliminate some observation errors and improve the fusion quality of the data.

[0022] For the multi-source fire point detection results in the overlapping area, extract the scale difference information from the multi-source fire point data set, and conduct a quantitative analysis on the extracted scale difference information to comprehensively describe the observation difference characteristics of fire points at different resolutions; For the multi-source fire point detection results in the overlapping area, extract the scale difference information from the multi-source fire point data set. The extracted information includes the number of small fire points split from a single low-resolution fire point at high resolution and the difference in pixel area variance within the fire point area at different resolutions. Conduct a quantitative analysis on the number of small fire points split from a single low-resolution fire point at high resolution and the difference in pixel area variance within the fire point area at different resolutions, and respectively generate a reference value for the fire point split number and a reference value for the local area variance. Comprehensively describe the observation difference characteristics of fire points at different resolutions through the reference value for the fire point split number and the reference value for the local area variance.

[0023] When a single low-resolution fire point splits into multiple small fire points in the high-resolution image and the split number increases abnormally, it usually indicates that there is an obvious scale error risk in the current multi-source fire point data. The essential reason for this phenomenon is that due to the large spatial sampling granularity of low-resolution remote sensing images, multiple adjacent or boundary-blurred fire points are often merged into a single overall fire point area. In high-resolution images, the original fire point boundary details are significantly magnified. Without scale consistency correction, during the fusion process, this low-resolution fire point will be wrongly decomposed into multiple independent high-resolution small fire points, forming the "fire point split" phenomenon. This phenomenon is not a real fire point distribution characteristic but is caused by the superposition of resolution differences and the amplification effect of scale differences, directly leading to problems such as distortion in fire point number statistics, disruption of fire point distribution continuity, and misjudgment of fire range and spread direction, seriously affecting the spatial accuracy and stability of fire monitoring. Therefore, the abnormal increase in the number of split fire points can be used as a significant indicator to measure the scale error risk in multi-source fire point data and can effectively reflect whether there is a fire point scale mismatch problem in the current fusion result.

[0024] The specific steps for quantitatively analyzing the number of small fire points split from a single low-resolution fire point at high resolution to generate a reference value for the fire point split number are as follows: Perform spatial matching on the low-resolution fire point areas and the fire point sets in the high-resolution images to judge the spatial coverage and inclusion relationships. For each low-resolution fire point, search and identify all the included high-resolution fire points within the corresponding spatial range to obtain the corresponding number of high-resolution split fire points. The calculation expression is as follows: ; In the formula, is the number of fire point splits, indicating the number of effective high-resolution fire points into which the low-resolution fire point is split under the high-resolution image, is the th high-resolution fire point, is the low-resolution fire point area, is the number of fire points in the high-resolution fire point candidate set, is the spatial inclusion indicator function, a function for judging whether the high-resolution fire point belongs to the range of the low-resolution fire point ; The function of the above steps is to accurately extract the number of "split fire points" in the high-resolution fire point set that are spatially related to the low-resolution fire points, and eliminate other fire points in the high-resolution image that are unrelated, ensuring the authenticity of the split number.

[0025] In order to characterize the severity of the split situation, combined with the scale characteristics of the low-resolution itself, a reference value for the fire point split number is generated to measure the relationship between the number of split fire points and the low-resolution spatial scale. The generation formula is as follows: ; In the formula, is the reference value for the fire point split number, is the boundary length (perimeter) of the low-resolution fire point, is the area of the low-resolution fire point.

[0026] By constructing the reference value for the fire point split number, the split degree of high- and low-resolution fire points in the spatial scale is quantitatively characterized, accurately reflecting the potential scale mismatch risk in the process of fusing multi-source fire point data. This step can explicitly identify potential fire point scale anomalies before fusion, providing a reliable risk criterion for subsequent adaptive correction.

[0027] The larger the reference value of the fire point split number generated after quantitatively analyzing the number of small fire points into which a single low-resolution fire point extracted splits at high resolution, generally means that there is a relatively high risk of scale error in the current multi-source fire point data. This is because, under normal circumstances, although there will be certain boundary refinement and morphological changes of the fire points at different resolutions, their overall spatial distribution and topological relationship should remain relatively consistent, and the number of split fire points should fluctuate within a reasonable range; if the split number increases abnormally, it indicates that a single fire point at low resolution is artificially split into multiple fire points at high resolution, which is obviously a scale mismatch phenomenon caused by the failure to fully adaptively adjust the resolution difference. At this time, the number, boundary shape and distribution pattern of the fire points in the fusion result will not conform to the actual fire situation, directly affecting the accuracy of fire monitoring and diffusion prediction. Therefore, the larger the reference value of the fire point split number, the higher the scale error risk; conversely, if the reference value of the fire point split number is small, it means that the fire point boundaries and scales at high and low resolutions are relatively coordinated, and generally it can be determined that there is no obvious scale error risk in the current multi-source fire point data.

[0028] The abnormal increase in the pixel area variance within the fire point area at different resolutions usually indicates that there is an obvious scale error risk in the current multi-source fire point data. The essential reason is that the scale error destroys the structural consistency of the fire points in space. Under normal circumstances, the pixel areas within the fire point areas in the low-resolution and high-resolution images should show relatively stable or interpretable changes, conforming to the proportional relationship of the image resolution scaling. However, if the variance of the fire point pixel area increases abnormally at high resolution, it means that the originally relatively uniform fire point area at low resolution is over-split, the boundary is broken or locally abnormally stretched in the high-resolution image, resulting in a highly uneven distribution of pixel areas within the area, thus causing local scale imbalance. This phenomenon generally occurs when there is a lack of adaptive scale correction in multi-source fire point fusion, especially more prominent in scenarios with complex terrain, boundaries between landforms or severe occlusion of low-resolution fire points, directly reflecting the inconsistency and fusion deviation of the current multi-source fire point observations in the spatial scale, which is a typical manifestation of scale error risk.

[0029] The specific steps for quantitatively analyzing the difference in the pixel area variance within the extracted fire point area at different resolutions to generate the local area variance reference value are as follows: For the key fire point areas in the high-resolution image and the low-resolution image, respectively extract the set of pixel areas within the area, denoted as and , is the set of pixel areas under the high-resolution image, is the set of pixel areas under the low-resolution image, calculate the area distribution deviation value, and the calculation expression is as follows: ; In the formula, is the area distribution deviation value, and are respectively the maximum and minimum values of the pixel area in the hotspot area at high resolution, and are respectively the maximum and minimum values of the pixel area in the hotspot area at low resolution, is a very small positive value to prevent the denominator from being zero (e.g., ); By the extreme value ratio of the pixel area sets of high-resolution and low-resolution images, it measures whether the dispersion degree of the area distribution at high resolution increases significantly relative to that at low resolution. The larger the area distribution deviation value, the more serious the non-uniform expansion or contraction of the local hotspot area in the high-resolution image, indicating potential scale distortion.

[0030] Calculate the fractional spacing deviation of the pixel area within the hotspot area, construct the local area variance reference value, and the calculation expression is as follows: ; In the formula, is the local area variance reference value, and are respectively the 75th and 25th percentiles of the pixel area at high resolution, represents the "upper layer" scale feature in the hotspot pixel area distribution at high resolution scale, which is used to measure the typical value of the larger pixel area in the hotspot and has stability, represents the "lower layer" scale feature in the hotspot pixel area distribution at high resolution scale, and together with forms the quantile difference of the high-resolution image, which can describe the dispersion of the area distribution, and are respectively the 75th and 25th percentiles of the pixel area at low resolution, reflects the typical value of the larger pixel area in the hotspot in the low-resolution image, is used as a reference index for high-low resolution comparison, and together with characterizes the inherent dispersion of the pixel area distribution in the hotspot area in the low-resolution image.

[0031] Characterize whether there is abnormal expansion or contraction of the local area distribution in the high-resolution image through the area distribution deviation degree at high and low resolutions; the quantile difference interval ( ) can be more sensitive to the outlier fluctuations of the local area; at the same time, is introduced as a magnification factor. Only when the extreme value deviation and the quantile difference are abnormal at the same time, the index value will be significantly magnified to reduce misjudgment; The larger the

[0032] The larger the local area variance reference value generated after quantitatively analyzing the differences in the pixel area variance within the extracted fire point regions at different resolutions, the more it usually indicates that the fire points in this region show a phenomenon of highly uneven pixel area distribution and abnormal local scale changes in high-resolution images. This situation is mostly caused by the merging, blurring or local occlusion of fire points at low resolutions, and the boundary fragmentation, fire point splitting, and scale mismatch at high resolutions, thereby reflecting a significant scale error risk in multi-source fire point data. On the contrary, when the local area variance reference value is small, that is, the pixel area variance within the fire point region changes smoothly and conforms to the normal proportional relationship under the resolution difference, it can usually be regarded as the multi-source fire points maintaining good stability and consistency in the spatial scale, indicating that there is no obvious scale error risk in the current multi-source fire point data.

[0033] Taking the quantitatively analyzed scale difference information as a feature parameter, input it into a pre-trained machine learning model, and evaluate the scale error risk of the current multi-source fire point data through the machine learning model; Taking the quantitatively analyzed fire point splitting number reference value and local area variance reference value as feature parameters, input them into a pre-trained machine learning model, generate a scale difference coefficient through the machine learning model, and evaluate the scale error risk of the current multi-source fire point data through the scale difference coefficient.

[0034] The "pre-trained machine learning model" refers to a fire point scale error risk assessment model obtained by training in a supervised learning manner in advance based on a large amount of historical remote sensing fire point data, real fire cases, and multi-source satellite observation data. In the training stage, representative historical multi-source fire point detection data are selected. For the known real fire point distribution, standard "scale error-free" fire point samples are constructed through manual annotation, measured data, or reliable high-resolution fire point extraction results. For each sample, multi-dimensional scale difference feature parameters including fire point splitting number, local area variance change, boundary curvature, main axis length ratio, fire point shape similarity, etc. are extracted, and a corresponding relationship is established with the actual scale error situation. Then, models such as support vector machine (SVM), random forest (RF), neural network (NN), or ensemble learning are used for training. By continuously inputting labeled data and scale features, the model automatically learns, fits, and masters the laws of scale distortion of multi-source fire points in different scenarios, and obtains the ability to predict risks for new samples. The establishment process of this model is to form a risk discrimination function by learning the "mapping relationship between resolution difference and fire point distortion risk", providing a data-driven determination basis for subsequent applications.

[0035] In the actual application stage, when faced with new multi-source hotspot data to be fused, the system first extracts the scale difference features of the current multi-source data (such as the reference value of hotspot splitting number and the reference value of local area variance) according to the aforementioned process, and inputs these feature parameters into the "pre-trained machine learning model". Since this model has learned the typical feature laws of resolution differences on hotspot scale distortion from a large amount of historical data, it can quickly analyze and discriminate the input features, and then output a scale difference coefficient that comprehensively reflects the scale distortion degree of the current data. The magnitude of the scale difference coefficient reflects whether there is an abnormal scale distortion risk in the fused hotspot. If the coefficient value is high, it indicates that there is a high-risk scale anomaly (such as boundary fragmentation, area distortion, hotspot splitting, etc.) in the current hotspot during the fusion process, and the system can promptly initiate subsequent scale consistency correction and boundary reconstruction strategies to actively eliminate potential scale deviation problems; if the coefficient value is low, it can directly enter the fusion process. By introducing the pre-trained machine learning model, not only can the scale anomaly discrimination problem of multi-source hotspot data under different terrains, resolutions, and hotspot types be effectively solved, the intelligence and accuracy of scale difference risk discrimination be improved, but also a highly reliable decision-making basis for subsequent dynamic adaptive fusion be provided.

[0036] The machine learning model is not limited here, and it can realize the reference value of hotspot splitting number and the reference value of local area variance to conduct comprehensive analysis and generate a scale difference coefficient Any machine learning model can be used. To implement the technical solution of the present invention, the present invention provides a specific implementation method; The formula for generating the scale difference coefficient is as follows: , where and are respectively the preset proportionality coefficients of the reference value of hotspot splitting number and the reference value of local area variance , and and are both greater than 0.

[0037] The "preset proportionality coefficient" refers to the coefficient that is determined artificially in advance or according to historical experience and model training results when calculating the scale difference coefficient to balance the relative importance of the reference value of hotspot splitting number and the reference value of local area variance in the comprehensive evaluation of scale difference. That is, the coefficients and respectively represent the weighted ratios of the reference value of hotspot splitting number and the reference value of local area variance when jointly generating the scale difference coefficient SDC.

[0038] Due to the reference value of the fire point splitting number and the reference value of the local area variance respectively reflect the fire point splitting characteristics and the fire point area change characteristics. Their sensitivities and action intensities to the scale error risk may be different in different terrains, seasons, and fire types. In order to make the generated more in line with the actual fire point scale difference risk characteristics, it is necessary to reasonably weight them through and when fusing these two characteristics. This ratio is not calculated from real-time observations, but is determined in advance according to training, experiments, or experience (i.e., "preset"). Generally, a relatively stable ratio can be obtained by fitting a large amount of historical data to ensure the balance of the contribution degrees of the two characteristics and meet the actual monitoring requirements. At the same time, it is stipulated that and must be greater than 0 to ensure that each characteristic plays a positive contribution in generation and avoid abnormal situations where the weights are 0 or negative.

[0039] Simply put: The preset proportional coefficient tells the model what proportions the "splitting characteristics" and "area characteristics" account for in the overall scale difference risk assessment.

[0040] From the scale difference coefficient, the larger the reference value of the fire point splitting number generated after quantitatively analyzing the number of small fire points split from a single low-resolution fire point extracted at high resolution, and the larger the reference value of the local area variance generated after quantitatively analyzing the differences in pixel area variance within the extracted fire point area at different resolutions, the larger the scale difference coefficient generated when evaluating the scale error risk of the current multi-source fire point data through a pre-trained machine learning model, indicating that the probability of the current multi-source fire point data having a scale error risk is greater. On the contrary, it indicates that the probability of the current multi-source fire point data having a scale error risk is smaller.

[0041] Compare and analyze the scale difference coefficient generated when evaluating the scale error risk of the current multi-source fire point data through a pre-trained machine learning model with the preset scale difference coefficient reference threshold to evaluate whether the current multi-source fire point data has a scale error risk. The specific steps are as follows: If the scale difference coefficient is greater than the preset scale difference coefficient reference threshold, it is determined that the current multi-source fire point data has a scale error risk; If the scale difference coefficient is less than or equal to the preset scale difference coefficient reference threshold, it is determined that the current multi-source fire point data does not have a scale error risk.

[0042] When the model identifies the risk of scale error, using the spatial scale and boundary features of high-resolution satellite data as the benchmark, a scale consistency correction algorithm is adopted to perform scale mapping and boundary reconstruction on the fire point detection results of low-resolution satellites. During the multi-source fire point fusion process, the fire point area boundary and spatial distribution characteristics are dynamically corrected to avoid the phenomenon of over-exaggeration or refinement of the fire point area after fusion. When the model identifies the risk of scale error, using the spatial scale and boundary features of high-resolution satellite data as the benchmark, a scale consistency correction algorithm is adopted to perform scale mapping and boundary reconstruction on the fire point detection results of low-resolution satellites. Its core function is to solve the problems of fire point boundary distortion, area distortion, and inconsistent spatial distribution caused by differences in spatial resolution during the multi-source remote sensing image fusion process. Since low-resolution satellites can often only roughly describe the spatial location and approximate range of fire points in fire point detection, while high-resolution satellites can provide more detailed fire point boundaries, shapes, and spatial distribution characteristics, this step uses high-resolution images as the correction benchmark, which can effectively utilize the more real and refined spatial scale information in high-resolution images to perform spatial scale mapping and boundary reconstruction on the fire point results detected by low-resolution satellites. Specifically, through the scale consistency correction algorithm, it is possible to adjust the characteristics of the fire point's shape, boundary position, area ratio, etc. one by one, correct the abnormal phenomena such as blurred boundaries, area scaling deviation, fire point splitting, and expansion caused by low-resolution observations, and then achieve dynamic optimization of the boundary and adaptive correction of the spatial distribution during the fire point fusion process. This step can not only effectively reduce problems such as false fire points, missed fire points, or abnormally refined and fragmented fire point boundaries caused by scale mismatch after the fusion of high and low-resolution fire points, but also improve the spatial consistency, scale reliability, and fire point characterization accuracy of the multi-source data fusion results, providing more reliable basic data for subsequent fire situation analysis, spread trend prediction, and emergency command and decision-making. This step is particularly crucial in the multi-source heterogeneous remote sensing data fusion task and is the core link to ensure the credibility, practicality, and engineering applicability of the fusion results.

[0043] When the model identifies the risk of scale error, using the spatial scale and boundary features of high-resolution satellite data as the benchmark, perform scale mapping and boundary reconstruction on the fire point detection results of low-resolution satellites. The specific steps for dynamically correcting the fire point area boundary and spatial distribution characteristics are as follows: In the multi-source fire point data set, using the fire point area of the low-resolution satellite as the reference area, extract the fire point boundary curvature distribution and area scale deformation factor within the area based on the high-resolution image, and construct a scale difference field. The formula is as follows: ; In the formula, is the difference intensity of the scale difference field at position , is the local curvature of the hotspot boundary in the high - resolution image at the position ; is the local curvature of the hotspot boundary in the low - resolution image at the corresponding position, is the local area scale function of the hotspot in the high - resolution image, is the local area scale function of the hotspot in the low - resolution image, is the curvature balance factor, which controls the relative contribution of the high - and low - resolution images in the curvature difference calculation, is the area scale difference weight factor, which controls the weight of the area scale difference in the overall scale difference field; The above steps establish a difference field of the high - and low - resolution hotspots in terms of local boundary morphology and area scale. Different from the simple area ratio or boundary distance measurement, it directly reflects the deformation - sensitive area of the hotspot and provides spatial guidance for subsequent correction.

[0044] Based on the difference intensity of the scale difference field , a scale - consistency vector field is constructed to guide the scale - consistency mapping of the low - resolution hotspot boundary points to the high - resolution reference boundary. The formula is as follows: ; In the formula, is the scale - consistency mapping vector field, which describes how the boundary points of the low - resolution hotspot at the position should be displaced to achieve the vector for scale - consistency correction, is the gradient of the difference intensity of the scale difference field, is the normal vector of the high - resolution hotspot boundary, which is the normal vector of the high - resolution hotspot boundary at the position , representing the outward normal direction of the boundary, is the gradient weight factor of the difference field, which controls the action intensity of the scale difference field gradient in the vector field, is the weight factor of the high - resolution boundary normal component, which controls the action intensity of the high - resolution hotspot boundary normal component in the vector field; The above steps do not directly perform simple contraction of the boundary points, but establish a scale - consistency vector field in the high - and low - resolution space through the joint drive of the difference field gradient + high - resolution boundary normal. This vector field guides the mapping of the low - resolution hotspot boundary to the best - matching direction of the high - resolution hotspot boundary, and at the same time automatically adapts to local scale differences, with high adaptability.

[0045] Using the scale - consistency mapping vector field to adjust the positions of the low - resolution hotspot boundary points, complete the boundary reconstruction, and complete the fusion of multi - source hotspots under the high - resolution reference. The calculation expression is as follows: ; Wherein, is the corrected boundary point position, representing the new position of the low-resolution fire point boundary point under the guidance of the scale consistency vector field and step size regulation, forming the reconstructed fire point boundary point. is the original position of the low-resolution fire point boundary point, representing the spatial coordinates of the fire point boundary point obtained after fire point detection and extraction in the low-resolution satellite image. is the step size adjustment factor, which controls the action intensity of the vector field on the boundary point, that is, the amplitude of the actual displacement of the boundary point under vector guidance.

[0046] The above steps complete the scale consistency reconstruction of the fire point boundary through continuous, local, and differential-driven vector displacements of the low-resolution fire point boundary, enabling its scale morphology to be adaptively corrected under the high-resolution benchmark, and avoiding the problems of "magnification" or "refinement" distortion of fire points under traditional resampling. After the reconstruction is completed, the multi-source fire points are fused uniformly under the high-resolution grid to obtain an accurate and scale-consistent fire spatial distribution.

[0047] The present invention effectively improves the spatial scale consistency and fusion accuracy of multi-source satellite fire point detection results, significantly improves problems such as fire point boundary fragmentation, false splitting, and abnormal area scaling caused by resolution differences, and ensures the accuracy and stability of the fused fire points in terms of spatial distribution, morphological characteristics, and scale information. Specifically, this solution dynamically extracts and quantifies scale difference information, combines a machine learning model to intelligently evaluate the scale error risk, and implements scale consistency correction based on high-resolution data as a benchmark, realizing the adaptive dynamic adjustment of the multi-source fire point boundary, effectively solving the problem of the lack of scale adaptability in multi-source fire point data fusion in the prior art, and improving the reliability and practicality of fire point detection results in fire monitoring, spread prediction, and emergency decision-making.

[0048] The above formulas are all dimensionless and take their numerical calculations. The formulas are obtained by collecting a large amount of data for software simulation to obtain a formula closest to the actual situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0049] Only some exemplary embodiments of the present invention have been described above by way of illustration. Undoubtedly, for those of ordinary skill in the art, without departing from the spirit and scope of the present invention, the described embodiments can be modified in various different ways. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the protection scope of the claims of the present invention.

[0050] It should be noted that in this text, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprising", "including" or any other variant thereof are intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or device comprising the element.

[0051] It should be understood that in various embodiments of the present application, the magnitude of the sequence numbers of the above processes does not imply the order of execution, and the order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0052] Those of ordinary skill in the art can realize that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods for each specific application to implement the described functions, but such implementation should not be considered to exceed the scope of the present application.

[0053] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0054] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0055] In addition, in each embodiment of the present application, the functional units can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit.

[0056] As described above, it is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, which should all be covered within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claimed rights.

[0057] Only some exemplary embodiments of the present invention have been described above by way of illustration. Without doubt, for those of ordinary skill in the art, the described embodiments can be modified in various different ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the protection scope of the claims of the present invention.

Claims

1. A method for fire point adaptation based on multi-source synchronous satellite data, characterized in that It includes the following steps: For satellite remote sensing images with different resolutions, fire point detection algorithms matching their resolution characteristics are respectively selected to independently complete the preliminary fire point detection, and the fire point spatial distribution and characteristic data corresponding to their respective resolution scales are extracted; The extracted multi-source fire point spatial distribution and characteristic data are preprocessed, and a multi-source fire point data set is established; For the multi-source fire point detection results in the overlapping area, scale difference information is extracted from the multi-source fire point data set, and the extracted scale difference information is quantitatively analyzed to comprehensively describe the observation difference characteristics of fire points at different resolutions; The quantified scale difference information is used as a characteristic parameter and input into a pre-trained machine learning model to evaluate the scale error risk of the current multi-source fire point data through the machine learning model; When the model identifies the existence of scale error risk, taking the spatial scale and boundary characteristics of high-resolution satellite data as the benchmark, a scale consistency correction algorithm is adopted to perform scale mapping and boundary reconstruction on the low-resolution satellite fire point detection results. During the multi-source fire point fusion process, the fire point area boundary and spatial distribution characteristics are dynamically corrected to avoid the phenomenon of over-exaggeration or refinement of the fire point area after fusion.

2. The fire point adaptive method based on multi-source synchronous satellite data according to claim 1, wherein The extracted multi-source fire point spatial distribution and characteristic data are preprocessed, and a multi-source fire point data set is established, which specifically includes the following steps: First, time synchronization processing is performed on the fire point spatial distribution and characteristic data from different satellites. Through observation time correction and interpolation completion, the time series error caused by inconsistent multi-source satellite observation times is solved; Secondly, unified spatial projection processing is carried out to uniformly convert each source data into the same geographic coordinate system and spatial resolution grid to ensure the corresponding relationship of each fire point position in space; Thirdly, noise elimination, outlier detection and correction are performed on the fire point characteristic data to eliminate false fire points or distorted characteristics caused by sensor errors and cloud and smoke interference; Then, for high-resolution, medium-resolution, and low-resolution satellite data, hierarchical storage and organization are carried out according to data source, resolution, and characteristic type to establish a structured multi-source fire point data set; Finally, it provides a standardized, spatio-temporally aligned, and feature-complete fire point observation data basis for subsequent scale difference analysis and fusion processing.

3. The fire point adaptive method based on multi-source synchronous satellite data according to claim 1, wherein For the multi-source fire point detection results in the overlapping area, scale difference information is extracted from the multi-source fire point data set. The extracted information includes the number of small fire points split from a single low-resolution fire point at high resolution and the difference in pixel area variance within the fire point area at different resolutions. The number of small fire points split from a single low-resolution fire point at high resolution and the difference in pixel area variance within the fire point area at different resolutions are quantitatively analyzed to respectively generate a reference value for the fire point split number and a reference value for the local area variance. The observation difference characteristics of fire points at different resolutions are comprehensively described through the reference value for the fire point split number and the reference value for the local area variance.

4. The fire point adaptive method based on multi-source synchronous satellite data according to claim 3, characterized in that The specific steps for quantitatively analyzing the number of small fire points split from a single low-resolution fire point at high resolution to generate a reference value for the fire point split number are as follows: Perform spatial matching on the low-resolution fire point areas and the fire point sets in the high-resolution images to determine the spatial coverage and inclusion relationships. For each low-resolution fire point, search and identify all the included high-resolution fire points within the corresponding spatial range to obtain the corresponding number of high-resolution split fire points. The calculation expression is as follows: ； In the formula, is the number of fire point splits, representing the number of low-resolution fire points split into the number of effective high-resolution fire points in the high-resolution image, is the th high-resolution fire point, is the low-resolution fire point area, is the number of fire points in the high-resolution fire point candidate set, is the spatial inclusion indicator function; In order to characterize the severity of the splitting situation, combined with the scale characteristics of the low-resolution itself, generate a reference value for the fire point split number to measure the relationship between the number of split fire points and the low-resolution spatial scale. The generation formula is as follows: ； In the formula, is the reference value of the fire point splitting number, is the boundary length of the low-resolution fire point, is the area of the low-resolution fire point.

5. The fire point adaptive method based on multi-source synchronous satellite data according to claim 3, wherein, The specific steps for quantitatively analyzing the differences in the pixel area variances within the extracted fire point areas at different resolutions to generate the local area variance reference value are as follows: For the key ignition point areas of high-resolution images and low-resolution images, the pixel area sets within the areas are extracted respectively, denoted as and , is the pixel area set under high-resolution images, is the pixel area set under low-resolution images. Calculate the area distribution deviation value, and the calculation expression is as follows: ； Wherein, is the area distribution deviation value, and are respectively the maximum and minimum values of the pixel area in the hotspot region at high resolution, and are respectively the maximum and minimum values of the pixel area in the hotspot region at low resolution, is a very small positive value to prevent the denominator from being zero; Calculate the fractional spacing deviation of the pixel areas within the fire point area and construct the local area variance reference value. The calculation expression is as follows: ； In the formula, is the reference value of the local area variance, and are the 75th and 25th percentiles of the pixel area at high resolution, respectively, and are the 75th and 25th percentiles of the pixel area at low resolution, respectively.

6. The fire point adaptive method based on multi-source synchronous satellite data according to claim 3, wherein Use the quantified fire point split number reference value and the local area variance reference value as feature parameters and input them into a pre-trained machine learning model. Generate a scale difference coefficient through the machine learning model and evaluate the scale error risk of the current multi-source fire point data through the scale difference coefficient.

7. The fire point adaptive method based on multi-source synchronous satellite data according to claim 1, characterized in that Compare and analyze the scale difference coefficient generated when evaluating the scale error risk of the current multi-source fire point data through the pre-trained machine learning model with the pre-set scale difference coefficient reference threshold to evaluate whether there is a scale error risk in the current multi-source fire point data. The specific steps are as follows: If the scale difference coefficient is greater than the pre-set scale difference coefficient reference threshold, it is determined that there is a scale error risk in the current multi-source fire point data; If the scale difference coefficient is less than or equal to the pre-set scale difference coefficient reference threshold, it is determined that there is no scale error risk in the current multi-source fire point data.

8. The adaptive fire point method based on multi-source satellite data according to claim 1, characterized in that When the model identifies a scale error risk, using the spatial scale and boundary characteristics of the high-resolution satellite data as a reference, perform scale mapping and boundary reconstruction on the low-resolution satellite fire point detection results, and dynamically correct the fire point area boundary and spatial distribution characteristics. The specific steps are as follows: In the multi-source fire point data set, using the fire point area of the low-resolution satellite as the reference area, extract the fire point boundary curvature distribution and area scale deformation factor within the area based on the high-resolution image to construct a scale difference field. The formula is as follows: ； In the formula, is the difference intensity of the scale difference field at the position , is the local curvature of the fire point boundary in the high-resolution image at the position , is the local curvature of the fire point boundary in the low-resolution image at the corresponding position, is the local area scale function of the fire point in the high-resolution image, is the local area scale function of the fire point in the low-resolution image, is the curvature balance factor, is the area scale difference weight factor; Based on the difference intensity of the scale difference field , a scale consistency vector field is constructed to guide the scale consistency mapping of the low-resolution fire point boundary points to the high-resolution reference boundary. The formula is as follows: ； In the formula, is the scale consistency mapping vector field, is the scale difference field difference intensity gradient, is the high-resolution fire point boundary normal vector, is the difference field gradient weight factor, is the high-resolution boundary normal component weight factor; Using Scale Consistency Mapping Vector Field Adjust the positions of the boundary points of the low-resolution fire points, complete the boundary reconstruction, and complete the fusion of multi-source fire points under the high-resolution benchmark. The calculation formula is as follows: ； In the formula, is the position of the corrected boundary point, is the original position of the boundary point of the low-resolution fire point, is the step size adjustment factor.

Images