Forest fire-fighting environment information acquisition method based on unmanned aerial vehicle
Patent Information
- Application Number
- CN202511017909.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-23
- Publication Date
- 2025-09-23
Smart Images
Figure CN120679105A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of forest fire source monitoring, and in particular to a method for collecting forest fire environment information based on an unmanned aerial vehicle (UAV). Background Art
[0002] To improve monitoring efficiency and timeliness, enhance monitoring precision and accuracy, and facilitate scientific firefighting command and decision-making, forest fire source monitoring can use drones equipped with sensors to collect information on the forest fire environment, ensuring continuous all-weather monitoring and detection data, and promoting the modernization of forest fire prevention management. When using drones to collect information on the firefighting environment at forest fire sites, infrared thermal imaging sensors are affected by ultra-high temperature heat sources in high-temperature fires. The radiation energy received by the photoelectric conversion unit will exceed the design threshold, causing the output signal to reach the upper limit and enter the nonlinear response range. At this time, the weak thermal radiation signal of the low-temperature residual fire is completely overwhelmed by the strong signal in the high-temperature area, forming a "white blind zone". In other words, the infrared thermal imaging sensor is prone to saturation in high-temperature fires, which also affects the identification of high-temperature heat sources and low-temperature residual fire areas.
[0003] To address this problem, the audio version uses adaptive threshold segmentation technologies such as the residual learning model based on the convolutional neural network (CNN) to identify artifact features in the saturated area through training data and reconstruct the masked low-temperature signal. The algorithm compensation effect is too dependent on the prior data set, and is prone to misjudgment when the fire scene morphology changes suddenly. For example, high-temperature fly ash is easily misidentified as residual fire, resulting in an increased false alarm rate. Summary of the Invention
[0004] The present invention provides a method for collecting forest fire environment information based on an unmanned aerial vehicle (UAV) to solve the problem that high-temperature signals saturate and mask low-temperature residual fire characteristics, easily misidentifying high-temperature fly ash as residual fire, and increasing the false alarm rate. The technical solution adopted is as follows:
[0005] An embodiment of the present invention provides a method for collecting forest fire environment information based on an unmanned aerial vehicle, the method comprising the following steps:
[0006] Collect the multispectral reflectance of the vegetation-covered area at the current collection time and at different previous collection times. At the same time, collect temperature data from the entire fire area and form a temperature distribution matrix. Delineate the unsaturated areas in the temperature distribution matrix. The temperature data is collected by an infrared thermal imaging sensor.
[0007] Using random forests, we classify each pixel in the unsaturated area as an afterfire based on the temperature data and multispectral reflectance data. We then determine the confidence level of each pixel in the unsaturated area. We then combine the differences between the temperature data of all pixels in the unsaturated area to determine the likelihood of an afterfire at the time the temperature data was collected, and establish a probability sequence for the existence of an afterfire at the current time of collection.
[0008] Perform time series trend analysis and clustering on the probability sequence of afterglow at the current collection time, and determine the afterglow diffusion index at the current collection time based on the time series trend analysis and clustering results;
[0009] According to the numerical relationship between the residual fire diffusion index at the current acquisition moment and the preset fire source monitoring threshold, it is determined whether the integration time of the infrared thermal imaging sensor should be adjusted. If so, the integration time of the infrared thermal imaging sensor is adjusted according to the residual fire diffusion index at the current acquisition moment. If not, no adjustment is made, thereby realizing accurate forest fire monitoring based on drones.
[0010] Furthermore, the classification confidence of the pixel belonging to the afterfire is determined by:
[0011] The temperature data and multispectral reflectance data of the unsaturated area are used as the input of the random forest. The trees corresponding to the pixel points are judged as belonging to the afterfire or not as the judgment results. The ratio of the number of trees belonging to the afterfire in the judgment results of each pixel point in the unsaturated area to the number of all trees in the judgment results of the pixel point is obtained and recorded as the classification confidence of the pixel point belonging to the afterfire.
[0012] Furthermore, the method for determining the possibility of the existence of afterglow at the time of collection corresponding to the temperature data is as follows:
[0013] According to the classification confidence of all pixels in the unsaturated area belonging to the afterfire, the cumulative classification confidence of the unsaturated area is determined;
[0014] The product of the cumulative classification confidence of the unsaturated area and the coefficient of variation of the temperature data of all pixels in the unsaturated area is recorded as the possibility of the existence of the afterfire at the collection time corresponding to the temperature data.
[0015] Furthermore, the cumulative classification confidence of the unsaturated area is determined based on the classification confidence of all pixels in the unsaturated area belonging to the afterfire, including the specific method of:
[0016] The cumulative sum of the classification confidences of all pixels in the unsaturated area belonging to afterfire is recorded as the cumulative classification confidence of the unsaturated area.
[0017] Furthermore, the specific method of establishing the possibility sequence of the existence of the afterglow at the current collection moment includes:
[0018] The afterglow existence probabilities at the current collection moment and all previous collection moments are arranged in the order of the collection moments to obtain an afterglow existence probability sequence at the current collection moment.
[0019] Furthermore, the afterfire existence possibility sequence at the current collection moment is subjected to time series trend analysis and clustering, and the afterfire diffusion index at the current collection moment is determined based on the time series trend analysis results and the clustering results, including the specific method of:
[0020] Performing a time series trend analysis on the possibility sequence of the afterglow at the current collection moment to obtain a trend coefficient of the possibility sequence of the afterglow at the current collection moment;
[0021] Perform density clustering on the possibility of afterfire existence at the current collection time and all previous collection times, and obtain the cluster radius of each cluster and the density of each core point of the cluster;
[0022] The afterfire diffusion index at the current collection moment is determined based on the trend coefficient of the afterfire existence possibility sequence at the current collection moment, the cluster radius of each cluster, and the density of each core point of the cluster.
[0023] Furthermore, the afterfire diffusion index at the current collection moment is determined based on the trend coefficient of the afterfire existence possibility sequence at the current collection moment, the cluster radius of each cluster, and the density of each core point of the cluster, including the specific method of:
[0024] The ratio of the cumulative sum of the densities of all core points in the same cluster to the cluster radius of the cluster is recorded as the first ratio of the cluster, and the cumulative sum of the first ratios of all clusters is recorded as the first cumulative sum at the current collection moment;
[0025] The product of the normalized value of the trend coefficient of the afterglow existence possibility sequence at the current collection moment and the first cumulative sum at the current collection moment is recorded as the first product at the current collection moment, and the normalized value of the first product at the current collection moment is recorded as the afterglow diffusion index at the current collection moment.
[0026] Furthermore, the method of determining whether to adjust the integration time of the infrared thermal imaging sensor based on the numerical relationship between the residual fire diffusion index at the current acquisition moment and the preset fire source monitoring threshold value includes the following specific methods:
[0027] If the residual fire diffusion index at the current collection moment is greater than the preset fire source monitoring threshold, the integration time of the infrared thermal imaging sensor is adjusted; if the residual fire diffusion index at the current collection moment is less than or equal to the preset fire source monitoring threshold, the integration time of the infrared thermal imaging sensor is not adjusted.
[0028] Furthermore, the specific method of adjusting the integration time of the infrared thermal imaging sensor according to the afterfire diffusion index at the current acquisition moment includes:
[0029] The hyperbolic tangent function value of the ratio of the afterfire diffusion index at the current collection moment to the preset fire source monitoring threshold is recorded as the first coefficient at the current collection moment;
[0030] The preset basic integration time is adjusted according to the first coefficient at the current acquisition moment to obtain an adjusted value of the integration time of the infrared thermal imaging sensor.
[0031] Furthermore, the preset basic integration time is adjusted according to the first coefficient at the current acquisition moment to obtain the adjusted value of the integration time of the infrared thermal imaging sensor, including the specific method of:
[0032] The product of the difference between the number 1 and the first coefficient at the current acquisition moment and the preset basic integration time is used as the adjustment value of the integration time of the infrared thermal imaging sensor.
[0033] The beneficial effects of the present invention are:
[0034] The present application divides the temperature detection area in which the sensor responds normally within the entire fire scene according to the collected temperature data, namely the unsaturated area. The unsaturated area is the core target area for afterfire fire monitoring. Afterfires or potential fire points that are not covered by high temperature can be effectively identified in the unsaturated area; the temperature fluctuation intensity and afterfire probability density in the unsaturated area are analyzed to evaluate the possibility of afterfire phenomenon in the unsaturated area, obtain the afterfire existence possibility at each collection moment, and construct the afterfire existence possibility sequence at the current collection moment; further, the afterfire existence possibility sequence at the current collection moment is subjected to time series trend analysis and clustering respectively, analyze the overall dynamic evolution trend of afterfire activity in the fire scene, the concentration and persistence of afterfire activity, and the local spatial scale of afterfire spread, evaluate the spatiotemporal coupling risk of afterfire activity, and obtain the afterfire diffusion index at the current collection moment. The larger the afterfire diffusion index at the current collection moment, the greater the possibility that afterfire activity gradually intensifies and a high-threat cluster exists, that is, the more likely the fire is to spread; finally, according to the afterfire at the current collection moment The fire spread index determines whether to adjust the integration time of the infrared thermal imaging sensor. If not, no adjustment is made. If so, the integration time of the infrared thermal imaging sensor is adjusted based on the afterfire spread index at the current acquisition moment. By shortening the integration time, the charge accumulation rate of the sensor chip is suppressed, preventing the photodiode in the high-temperature area from entering the nonlinear saturation region due to overexposure, thereby retaining the sensor's ability to detect weak afterfire signals in the non-saturated area, and realizing accurate forest fire monitoring based on drones. The adjusted integration time of the infrared thermal imaging sensor can adaptively suppress infrared saturation, while balancing the monitoring accuracy of low-temperature afterfires and adaptively suppressing signal saturation in high-temperature areas. It can effectively restore the weak thermal radiation characteristics of obscured low-temperature afterfires. While ensuring temperature accuracy in non-saturated areas, it reduces the false alarm rate when the fire scene morphology suddenly changes, and enhances the dynamic identification capability of hidden fire points. This can achieve a synergistic improvement in fire monitoring accuracy and robustness, and solve the problem of high-temperature signal saturation masking low-temperature afterfire characteristics, which easily misjudges high-temperature fly ash as afterfire, resulting in an increased false alarm rate. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0036] Figure 1 A flowchart of a method for collecting forest fire environment information based on a drone is provided in accordance with one embodiment of the present invention;
[0037] Figure 2A flowchart for obtaining the possibility of afterfire existence is provided in one embodiment of the present invention. DETAILED DESCRIPTION
[0038] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.
[0039] See also Figure 1 , which shows a flow chart of a method for collecting forest fire environment information based on a drone provided by an embodiment of the present invention, the method comprising the following steps:
[0040] Step S001: Collect the multispectral reflectance of the vegetation coverage area at the current collection time and different previous collection times, and at the same time collect temperature data across the entire fire scene and form a temperature distribution matrix, dividing the temperature distribution matrix into high-temperature saturated areas and unsaturated areas, wherein the temperature data is collected by an infrared thermal imaging sensor.
[0041] An infrared thermal imaging sensor is integrated under the drone fuselage, and multispectral imaging equipment is symmetrically mounted on the drone's wings. The infrared thermal imaging sensor is used to collect temperature radiation data across the fire scene, and the multispectral imaging equipment is used to obtain the multispectral reflectivity of vegetation-covered areas.
[0042] Temperature radiation data is used to accurately locate high-temperature fire sources and potential residual fire areas. Infrared thermal imaging sensors receive this data. Based on Planck's radiation law and sensor calibration parameters, the radiation intensity at each pixel within the fire area is converted into an absolute temperature value. This data is then mapped to spatial coordinates to generate a two-dimensional temperature matrix, providing real-time temperature distribution within the fire area. In other words, infrared thermal imaging sensors can collect temperature data for every pixel within the fire area, which is then combined into a temperature distribution matrix.
[0043] Multispectral reflectance includes infrared, near-infrared and visible light band reflectance data of vegetation coverage areas, which is used to invert vegetation moisture content, combustion status and combustible material distribution characteristics.
[0044] Preferably, in one embodiment of the present application, when collecting temperature radiation data and multispectral reflectance, the data sampling frequency in this embodiment is 10 Hz. In actual application, as other implementation methods, implementers can determine the sampling rate according to actual conditions, and this application does not impose any special restrictions.
[0045] At this point, the temperature data, temperature distribution matrix, and multispectral reflectance at the current sampling moment and each previous sampling moment are obtained.
[0046] To balance local feature sensitivity and computational efficiency, the analysis window for the local grayscale histogram was set to a 5×5 pixel window. Adaptive threshold segmentation was applied to the temperature distribution matrix to identify high-temperature saturated and unsaturated regions. The high-temperature saturated region represents areas of extreme temperature where the infrared thermal imaging sensor cannot accurately measure due to the received radiation energy exceeding the threshold. The unsaturated region represents the temperature detection range within the sensor's normal response range. This region effectively identifies embers or potential fires that are not obscured by high temperatures and is a key target area for ember fire monitoring. Adaptive threshold segmentation algorithms can employ methods such as adaptive threshold segmentation based on the local neighborhood mean and adaptive threshold segmentation based on Gaussian weights.
[0047] At this point, the high-temperature saturated area and non-saturated area, multispectral reflectance and temperature data in the temperature distribution matrix are obtained.
[0048] Step S002: Use random forest to classify whether each pixel in the unsaturated area is an afterfire, determine the classification confidence of the pixel as an afterfire, combine the differences between the temperature data of all pixels in the unsaturated area, determine the possibility of afterfire at the collection time corresponding to the temperature data, and establish the possibility sequence of afterfire at the current collection time.
[0049] The temperature data and multispectral reflectance data of the unsaturated area are used as input, and random forest is used to perform multi-feature fusion classification on whether each pixel point in the unsaturated area is an afterfire. The judgment results for each pixel point in the unsaturated area are the number of trees belonging to "afterfire", the number of all trees, and the classification confidence that the pixel point belongs to afterfire.
[0050] The number of decision trees is set to 100, Gini impurity is used as the splitting criterion, and the maximum depth of each tree is limited to 10 layers. Random forest fuses the temperature gradient of temperature data with the vegetation moisture, NDVI, and NDWI of multispectral data into feature vectors. The probability classification of afterfire is realized through the feature division of decision tree nodes. Random forest generates probabilities through a voting mechanism. All decision trees independently classify each pixel point. The results of independent classification are afterfire and non-afterfire. Finally, the output is the tree that supports "afterfire" and the tree that supports "non-afterfire" for each pixel point, as well as the classification confidence of the pixel point belonging to afterfire. The judgment result of the pixel point is the ratio of the number of trees belonging to "afterfire" to the number of all trees corresponding to the pixel point, that is, the classification confidence of the pixel point. NDVI is the Normalized Difference Vegetation Index, which is an indicator to measure the growth status and greenness of vegetation. NDWI is the Normalized Difference Vegetation Index, which is an indicator to measure the growth status and greenness of vegetation. WaterIndex, or the Normalized Difference Water Index, is a vegetation index used to identify water bodies. NDVI (NDWI) is determined based on multispectral reflectance. Using random forests to perform multi-feature fusion classification on each pixel in the unsaturated area is a well-known technique and will not be described in detail.
[0051] The possibility of the existence of afterglow at the collection time corresponding to the temperature data is determined based on the difference between the temperature data of all pixels in the unsaturated area and the classification confidence that all pixels in the unsaturated area belong to afterglow.
[0052] Preferably, as an embodiment of the present application, the cumulative sum of the classification confidences of all pixels in the unsaturated area belonging to afterfire is recorded as the cumulative classification confidence of the unsaturated area, and the product of the cumulative classification confidence of the unsaturated area and the coefficient of variation of the temperature data of all pixels in the unsaturated area is recorded as the possibility of the existence of afterfire at the collection time corresponding to the temperature data.
[0053] The coefficient of variation (Cov) is a statistic that measures the relative dispersion of data. It represents the ratio of the standard deviation to the mean and is typically expressed as a percentage. The calculation of the Cov of temperature data is well known and will not be further explained. The Cov of the temperature data for all pixels within an unsaturated region quantifies the degree of dispersion of the temperature data distribution within the unsaturated region. A larger Cov value indicates a more dispersed temperature distribution and a higher likelihood of the presence of dynamic afterfires within the unsaturated region, where the temperature fluctuates dramatically due to localized fires.
[0054] The classification confidence of the pixel points in the unsaturated area belonging to the afterfire indicates the significance of the pixel points showing features consistent with the afterfire. The cumulative sum of the classification confidences of all pixels in the unsaturated area belonging to the afterfire is the significance of the unsaturated area showing features consistent with the afterfire. When the value is larger, the spatial density of the afterfire points in the unsaturated area is greater.
[0055] When the possibility of afterfire at the time of acquisition is greater, the temperature fluctuation in the unsaturated area is more intense and the probability density of afterfire is higher. At this time, the possibility of afterfire in the unsaturated area is more likely to exist. Figure 2 shown.
[0056] The afterglow existence probabilities at the current collection moment and all previous collection moments are arranged in the order of the collection moments to obtain the afterglow existence probability sequence at the current collection moment.
[0057] It is understandable that, for the first collection moment, the corresponding afterglow existence possibility sequence only contains one value, and the first collection moment is not analyzed.
[0058] At this point, the ember existence possibility sequence at the current collection moment is obtained.
[0059] Step S003: performing time series trend analysis and clustering on the afterfire existence possibility sequence at the current collection moment, and determining the afterfire diffusion index at the current collection moment based on the time series trend analysis results and the clustering results.
[0060] Forest fires are characterized by dynamic diffusion and spatial heterogeneity. Traditional monitoring methods find it difficult to capture the spatiotemporal coupled evolution of fires in real time, which often leads to delayed identification of residual fires and inaccurate assessment of fire spread risks. Especially under the cover of high-temperature saturated areas, hidden fire points are easily missed or fly ash interference is misjudged.
[0061] Therefore, the ARIMA model is used to perform time series trend analysis on the probability sequence of the existence of the afterglow at the current collection time. The value of the autoregressive term is set to 3, the value of the difference order is set to 1, and the value of the sliding average term is set to 2 to obtain the trend coefficient of the probability sequence of the existence of the afterglow at the current collection time.
[0062] Among them, the trend coefficient of the afterfire possibility sequence represents the increasing or decreasing trend of the afterfire possibility over time, reflecting the intensity of the dynamic evolution law of the overall strengthening or weakening of afterfire activity in the unsaturated area.
[0063] The DBSCAN density clustering algorithm is used to cluster all the afterfire existence possibilities contained in the afterfire existence possibility sequence at the current collection moment. The neighborhood radius is set to 50 and the minimum density point is set to 5. The clustering radius of each cluster and the density of each core point of the cluster are obtained.
[0064] Clustering is performed using the DBSCAN density clustering algorithm. Obtaining the cluster radius and the density of each core point in each cluster are well-known techniques and will not be further described. The cluster radius reflects the local spatial scale of the dynamic ember spread within the fire scene, while the density of each core point in the cluster reflects the concentration and persistence of the ember activity.
[0065] The ratio of the cumulative sum of the densities of all core points in the same cluster to the cluster radius of the cluster is recorded as the first ratio of the cluster, and the cumulative sum of the first ratios of all clusters is recorded as the first cumulative sum at the current collection time. The product of the normalized value of the trend coefficient of the afterfire existence possibility sequence at the current collection time and the first cumulative sum at the current collection time is recorded as the first product at the current collection time, and the normalized value of the first product at the current collection time is recorded as the afterfire diffusion index at the current collection time.
[0066] It should be noted that the value range of the trend coefficient of the possibility sequence is greater than or equal to -1 and less than or equal to 1. This embodiment uses the Z-Score standard normalization method to calculate the normalized value. In actual application, the implementer can use other existing methods such as the maximum and minimum value normalization method, the sigmoid function, etc. to calculate the normalized value, which is not limited here.
[0067] The normalized value of the trend coefficient of the afterfire probability sequence at the current collection time reflects the overall dynamic evolution of afterfire activity within the fire scene. A larger value indicates a more pronounced trend of increasing afterfire activity over time, and a higher risk of fire spread or rekindling. The cumulative sum of the densities of all core points within a cluster reflects the concentration and persistence of afterfire activity. A larger cumulative sum indicates a higher density of afterfire points within the cluster, and a greater likelihood of high-intensity combustion or persistent rekindling in a localized area. The cluster radius reflects the local spatial scale of afterfire spread. A smaller cluster radius indicates a higher likelihood of afterfires concentrated in a confined area being high-intensity fire points, and a higher heat release rate. The first cumulative sum at the current collection time reflects the comprehensive spatial threat of afterfire activity within the fire scene. A larger first cumulative sum indicates a greater number of high-density, small-scale afterfire clusters, and a higher threat level. The afterfire diffusion index at the current collection moment comprehensively quantifies the spatiotemporal coupling risk of afterfire activity. The larger the afterfire diffusion index at the current collection moment, the greater the possibility that afterfire activity will gradually intensify and a high-threat cluster will exist, that is, the more likely the fire will spread.
[0068] At this point, the residual fire diffusion index at the current collection moment is obtained.
[0069] Step S004: Determine whether to adjust the integration time of the infrared thermal imaging sensor based on the numerical relationship between the residual fire diffusion index at the current acquisition moment and the preset fire source monitoring threshold. If so, adjust the integration time of the infrared thermal imaging sensor based on the residual fire diffusion index at the current acquisition moment; if not, do not adjust it, thereby realizing accurate forest fire monitoring based on drones.
[0070] The thermal radiation intensity of high-temperature areas in forest fires far exceeds the design threshold of infrared thermal imaging sensors, causing the photoelectric conversion unit to enter a nonlinear response range. The high-temperature saturation signal completely obscures the weak thermal radiation characteristics of low-temperature afterglow, forming a "white blind zone." Traditional compensation methods rely on static prior data, which can easily lead to misjudgment of forest fire source monitoring results when sudden changes in fire morphology occur, such as interference from high-temperature fly ash or rapid fire spread. This leads to increased false alarm rates and ineffective afterglow detection. Therefore, an improved method is established to establish a dynamic feedback mechanism based on the afterglow diffusion index at the current acquisition time. Specifically, the improved method is as follows.
[0071] Compare the afterfire spread index at the current acquisition time with the fire source monitoring threshold. If the afterfire spread index at the current acquisition time is greater than the fire source monitoring threshold, the infrared thermal imaging sensor's integration time needs to be adjusted. At this time, the spatiotemporal coupling of afterfire activity is enhanced, and the risk of fire spread and re-ignition increases dramatically. Adjusting the infrared thermal imaging sensor's integration time can reduce the exposure intensity in high-temperature areas, avoid signal overflow, and prevent signal saturation in high-temperature areas. If the afterfire spread index at the current acquisition time is less than or equal to the fire source monitoring threshold, the infrared thermal imaging sensor's integration time does not need to be adjusted.
[0072] The fire source monitoring threshold is a preset parameter value. In this embodiment, the fire source monitoring threshold is set to 0.6.
[0073] Among them, the calculation method of the adjusted integration time of the infrared thermal imaging sensor is: the hyperbolic tangent function value of the ratio of the residual fire diffusion index at the current acquisition moment to the fire source monitoring threshold is recorded as the first coefficient at the current acquisition moment, and the product of the difference between the number 1 and the first coefficient at the current acquisition moment and the basic integration time is recorded as the adjusted integration time of the infrared thermal imaging sensor.
[0074] The basic integration time is a preset time length, which is a theoretical optimal value set according to the maximum signal-to-noise ratio of the sensor. In this embodiment, the basic integration time is set to 5 ms.
[0075] When the afterfire diffusion index at the current acquisition moment is greater than the fire source monitoring threshold, the adjusted integration time of the infrared thermal imaging sensor is calculated and used as the value of the integration time. By shortening the integration time, the charge accumulation rate of the sensor chip is suppressed, and the photodiode in the high-temperature area is prevented from entering the nonlinear saturation region due to overexposure, thereby retaining the sensor's ability to detect weak afterfire signals in the non-saturated area.
[0076] The adjusted integration time of the infrared thermal imaging sensor adaptively suppresses infrared saturation through the fire source monitoring threshold and nonlinear adjustment method, while balancing the monitoring accuracy of low-temperature residual fires and adaptively suppressing the signal saturation phenomenon in high-temperature areas. It can effectively restore the weak thermal radiation characteristics of masked low-temperature residual fires, while ensuring the temperature accuracy of non-saturated areas, reducing the false alarm rate when the fire scene morphology changes suddenly, and enhancing the dynamic identification ability of hidden fire points, which can achieve a coordinated improvement in fire monitoring accuracy and robustness.
[0077] At this point, accurate monitoring of forest fires based on drones has been achieved.
[0078] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the principles of the present invention should be included in the scope of protection of the present invention.
Claims
1. A method for collecting forest fire environment information based on drones, characterized in that: The method comprises the following steps: Collect the multispectral reflectance of the vegetation-covered area at the current collection time and at different previous collection times. At the same time, collect temperature data from the entire fire area and form a temperature distribution matrix. Delineate the unsaturated areas in the temperature distribution matrix. The temperature data is collected by an infrared thermal imaging sensor. Using random forests, we classify each pixel in the unsaturated area as an afterfire based on the temperature data and multispectral reflectance data. We then determine the confidence level of each pixel in the unsaturated area. We then combine the differences between the temperature data of all pixels in the unsaturated area to determine the likelihood of an afterfire at the time the temperature data was collected, and establish a probability sequence for the existence of an afterfire at the current time of collection. Perform time series trend analysis and clustering on the probability sequence of afterglow at the current collection time, and determine the afterglow diffusion index at the current collection time based on the time series trend analysis and clustering results; According to the numerical relationship between the residual fire diffusion index at the current acquisition moment and the preset fire source monitoring threshold, it is determined whether the integration time of the infrared thermal imaging sensor should be adjusted. If so, the integration time of the infrared thermal imaging sensor is adjusted according to the residual fire diffusion index at the current acquisition moment. If not, no adjustment is made, thereby realizing accurate forest fire monitoring based on drones.
2. The method for collecting forest fire environment information based on an unmanned aerial vehicle according to claim 1, characterized in that: The method for determining the classification confidence of the pixel belonging to the afterfire is: The temperature data and multispectral reflectance data of the unsaturated area are used as the input of the random forest. The trees corresponding to the pixel points are judged as belonging to the afterfire or not as the judgment results. The ratio of the number of trees belonging to the afterfire in the judgment results of each pixel point in the unsaturated area to the number of all trees in the judgment results of the pixel point is obtained and recorded as the classification confidence of the pixel point belonging to the afterfire.
3. The method for collecting forest fire environment information based on an unmanned aerial vehicle according to claim 1, characterized in that: The method for determining the possibility of the existence of the afterglow at the time of collection corresponding to the temperature data is as follows: According to the classification confidence of all pixels in the unsaturated area belonging to the afterfire, the cumulative classification confidence of the unsaturated area is determined; The product of the cumulative classification confidence of the unsaturated area and the coefficient of variation of the temperature data of all pixels in the unsaturated area is recorded as the possibility of the existence of the afterfire at the collection time corresponding to the temperature data.
4. The method for collecting forest fire environment information based on an unmanned aerial vehicle according to claim 3, characterized in that: The method of determining the cumulative classification confidence of the unsaturated area according to the classification confidence of all pixels in the unsaturated area belonging to the afterfire includes the following specific methods: The cumulative sum of the classification confidences of all pixels in the unsaturated area belonging to afterfire is recorded as the cumulative classification confidence of the unsaturated area.
5. The method for collecting forest fire environment information based on an unmanned aerial vehicle according to claim 1, characterized in that: The specific method of establishing the possibility sequence of the existence of the embers at the current collection moment includes: The afterglow existence probabilities at the current collection moment and all previous collection moments are arranged in the order of the collection moments to obtain an afterglow existence probability sequence at the current collection moment.
6. The method for collecting forest fire environment information based on an unmanned aerial vehicle according to claim 1, characterized in that: The method of performing time series trend analysis and clustering on the afterfire existence possibility sequence at the current collection time, and determining the afterfire diffusion index at the current collection time according to the time series trend analysis results and the clustering results, includes the following specific methods: Performing a time series trend analysis on the possibility sequence of the afterglow at the current collection moment to obtain a trend coefficient of the possibility sequence of the afterglow at the current collection moment; Perform density clustering on the possibility of afterfire existence at the current collection time and all previous collection times, and obtain the cluster radius of each cluster and the density of each core point of the cluster; The afterfire diffusion index at the current collection moment is determined based on the trend coefficient of the afterfire existence possibility sequence at the current collection moment, the cluster radius of each cluster, and the density of each core point of the cluster.
7. The method for collecting forest fire environment information based on an unmanned aerial vehicle according to claim 6, characterized in that: The method of determining the afterfire diffusion index at the current collection moment based on the trend coefficient of the afterfire existence possibility sequence at the current collection moment, the cluster radius of each cluster, and the density of each core point of the cluster includes the following specific methods: The ratio of the cumulative sum of the densities of all core points in the same cluster to the cluster radius of the cluster is recorded as the first ratio of the cluster, and the cumulative sum of the first ratios of all clusters is recorded as the first cumulative sum at the current collection moment; The product of the normalized value of the trend coefficient of the afterglow existence possibility sequence at the current collection moment and the first cumulative sum at the current collection moment is recorded as the first product at the current collection moment, and the normalized value of the first product at the current collection moment is recorded as the afterglow diffusion index at the current collection moment.
8. The method for collecting forest fire environment information based on an unmanned aerial vehicle according to claim 1, characterized in that: The specific method for determining whether to adjust the integration time of the infrared thermal imaging sensor based on the numerical relationship between the residual fire diffusion index at the current acquisition moment and the preset fire source monitoring threshold is as follows: If the residual fire diffusion index at the current collection moment is greater than the preset fire source monitoring threshold, the integration time of the infrared thermal imaging sensor is adjusted; if the residual fire diffusion index at the current collection moment is less than or equal to the preset fire source monitoring threshold, the integration time of the infrared thermal imaging sensor is not adjusted.
9. The method for collecting forest fire environment information based on an unmanned aerial vehicle according to claim 8, characterized in that: The specific method of adjusting the integration time of the infrared thermal imaging sensor according to the afterfire diffusion index at the current acquisition moment includes: The hyperbolic tangent function value of the ratio of the afterfire diffusion index at the current collection moment to the preset fire source monitoring threshold is recorded as the first coefficient at the current collection moment; The preset basic integration time is adjusted according to the first coefficient at the current acquisition moment to obtain an adjusted value of the integration time of the infrared thermal imaging sensor.
10. The method for collecting forest fire environment information based on an unmanned aerial vehicle according to claim 9, characterized in that: The method of adjusting the preset basic integration time according to the first coefficient at the current acquisition moment to obtain the adjustment value of the integration time of the infrared thermal imaging sensor includes the following specific methods: The product of the difference between the number 1 and the first coefficient at the current acquisition moment and the preset basic integration time is used as the adjustment value of the integration time of the infrared thermal imaging sensor.