A method and system for early fire detection based on multispectral spectrum

Abstract

This invention provides a multispectral early fire point identification method and system. It involves multi-dimensional data acquisition of an early fire point monitoring area, obtaining a raw fire point monitoring dataset and performing data preprocessing to obtain a fire point monitoring dataset. Features are extracted from the fire point monitoring dataset to obtain a multi-dimensional fire point feature dataset. Early fire point identification is performed based on a multi-dimensional collaborative fire point identification strategy and the multi-dimensional fire point feature dataset, yielding early fire point identification results. By introducing multi-dimensional data for multi-dimensional analysis of early fire points, the accuracy of early fire point identification is improved. Based on the early fire point identification results, early fire point verification decisions are executed, and early fire point verification is performed based on these decisions, yielding early fire point verification results. Finally, the multi-dimensional collaborative fire point identification strategy is optimized based on the early fire point verification results, thereby improving the accuracy and stability of early fire point identification.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and more specifically, to a method and system for early fire point identification based on multispectral imaging. Background Technology

[0002] In the field of fire prevention and control, early fire point identification is an indispensable part of reducing disaster losses. Traditional solutions often rely on single spectral features or simple threshold judgments. However, traditional solutions often exhibit limitations in complex scenarios, such as weak anti-interference capabilities and insufficient early warning accuracy.

[0003] On the one hand, traditional solutions identify fire points through simple threshold judgment and single spectral features. However, the features corresponding to real fire points often cannot be accurately identified by analyzing single data. At the same time, although simple static preset threshold judgment can filter out a certain amount of interference source data, a large amount of interference source data is still not filtered out. This makes it difficult to identify and locate "real fire points" when distinguishing between "real fire points" and "fire point-like interference sources". On the other hand, in real environments, early fire point identification through single spectral features is easily affected by external factors, resulting in poor accuracy and stability of early fire point identification.

[0004] In summary, traditional solutions mostly rely on feature analysis of a single spectrum and simple judgment conditions such as static thresholds to identify early fire points, which often fails to meet the needs of accurate early fire point identification in complex scenarios. Therefore, it is urgent to construct a multispectral early fire point identification method to break through the inherent limitations of traditional technical solutions and improve the accuracy of early fire point identification. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a method and system for early fire point identification based on multispectral imaging. The objectives and effects of this method and system are achieved through the following specific technical means:

[0006] A method for early fire detection based on multispectral imaging includes:

[0007] Multi-dimensional data collection was conducted on the early fire point monitoring area to obtain the raw fire point monitoring dataset, and the raw fire point monitoring dataset was preprocessed to obtain the fire point monitoring dataset.

[0008] Feature extraction is performed on the fire point monitoring dataset to obtain a multidimensional fire point feature dataset, which includes ecological balance disturbance feature data, thermal convection motion feature data, and environmental element synergy feature data.

[0009] Early fire point identification is performed based on a multi-dimensional collaborative fire point identification strategy and a multi-dimensional fire point feature dataset to obtain early fire point identification results.

[0010] Based on the early fire point identification results, perform early fire point verification decisions and obtain the early fire point verification results;

[0011] The multi-dimensional collaborative fire point identification strategy was optimized based on the early fire point verification results.

[0012] As a further aspect of the present invention, feature extraction is performed on the fire point monitoring dataset to obtain a multi-dimensional fire point feature dataset, including:

[0013] The time series data of ambient temperature monitoring data contained in the fire point monitoring dataset are analyzed to obtain the results of dynamic time series temperature changes. Based on the results of dynamic time series temperature changes, the intensity of local thermal disturbance in the early fire point monitoring area is quantified, and thermal convection motion characteristic data are generated.

[0014] Based on the combustion gas concentration monitoring data, plant water loss monitoring data, and plant physiological activity monitoring data contained in the fire point monitoring dataset, an ecological balance disturbance correlation map is constructed, and ecological balance disturbance characteristic data is generated by analyzing the temporal interaction correlation of the ecological balance disturbance correlation map.

[0015] Based on the environmental causal coupling mechanism, the environmental disturbance data and plant water loss monitoring data contained in the fire point monitoring dataset are analyzed to generate environmental element synergistic feature data.

[0016] A multidimensional fire point feature dataset was constructed based on thermal convection motion characteristic data, ecological balance disturbance characteristic data, and environmental element collaborative characteristic data.

[0017] As a further aspect of the present invention, environmental disturbance data and plant water loss monitoring data contained in the fire point monitoring dataset are analyzed based on the environmental causal coupling mechanism to generate environmental element synergistic feature data, including:

[0018] By reconstructing and integrating time-series data of environmental disturbance data and plant water loss monitoring data in phase space, and generating corresponding phase space motion trajectories, chaotic feature data corresponding to the sudden changes in phase space motion trajectory caused by early fire points are obtained.

[0019] The causal strength between environmental disturbance data and plant water loss monitoring data is calculated based on the convergent cross-mapping algorithm. A causal feature network is constructed based on the causal transmission path and corresponding causal strength between the environmental disturbance data and plant water loss monitoring data. Causal feature data is obtained based on the causal feature network.

[0020] The fluctuation phase of environmental disturbance data and plant water loss monitoring data was extracted by Hilbert transform, and phase-coordinated feature data was obtained based on the fluctuation phase.

[0021] Chaotic feature data, causal feature data, and phase-coordinated feature data are encapsulated and output as environmental element coordinated feature data.

[0022] As a further aspect of the present invention, early fire point identification is performed based on a multi-dimensional collaborative fire point identification strategy and a multi-dimensional fire point feature dataset to obtain early fire point identification results, including:

[0023] The thermal convection motion characteristic data is decomposed into velocity field, acceleration field and vortex field, and the expansion mode feature is extracted by spatiotemporal convolutional network to obtain expansion mode feature data. Based on the expansion mode feature data, the fire point thermal plume index is generated. When the fire point thermal plume index is greater than the preset thermodynamic pressure threshold, it is determined that a thermodynamic anomaly has occurred in the early fire point monitoring area, and the spatial location of the thermodynamic anomaly corresponding to the early fire point monitoring area is obtained.

[0024] A multidimensional collaborative fire point correlation map is constructed based on the ecological balance disturbance characteristic data and environmental element synergistic characteristic data corresponding to the thermodynamic anomaly area. Fire point inference is performed based on the multidimensional collaborative fire point correlation map to obtain the fire point inference results.

[0025] Early fire point identification results are generated based on the fire point thermal plume index and fire point inference results.

[0026] As a further aspect of the present invention, an early fire point identification result is generated based on the fire point thermal plume index and fire point inference results, including:

[0027] Obtain the spatial location of the inference association region associated with the fire point inference result, and perform spatial overlap analysis on the spatial location of the thermodynamic anomaly and the inference association spatial location to obtain the spatial overlap.

[0028] When the spatial overlap is not lower than the preset overlap threshold, it is determined that an early fire point has occurred. The risk level is generated based on the fire point thermal plume index and the fire point inference result, and the spatial location and risk level corresponding to the early fire point are output.

[0029] When the spatial overlap is lower than the preset overlap threshold, it is determined that no early fire point has occurred.

[0030] As a further aspect of the present invention, performing an early fire point verification decision based on the early fire point identification result and obtaining the early fire point verification result includes:

[0031] Analyze the early fire point identification results to obtain the risk level of the early fire point;

[0032] The risk levels include high risk, medium risk, and low risk.

[0033] Decisions for early fire detection are made based on risk level:

[0034] For early fire point areas at high and medium risk levels, multi-source collaborative verification of early fire points is conducted.

[0035] For early fire zones with low risk levels, manual verification is conducted.

[0036] Early fire point verification results are obtained through early fire point verification decisions, and fire point elimination strategies are executed based on these results.

[0037] As a further aspect of the present invention, the multi-dimensional collaborative fire point identification strategy is optimized based on early fire point verification results, including:

[0038] The confidence level of the rules contained in the multi-dimensional collaborative fire point identification strategy corresponding to the successful early fire point identification in the early fire point verification results is enhanced.

[0039] For cases where early fire point identification fails in the early fire point verification results, interference source diagnosis is performed to obtain the interference source diagnosis results, and the interference source diagnosis results are expanded to the preset non-fire point interference source feature library.

[0040] As a further aspect of the present invention, multi-dimensional data collection is performed on the early fire point monitoring area to obtain a raw fire point monitoring dataset, and the raw fire point monitoring dataset is preprocessed to obtain a fire point monitoring dataset, including:

[0041] Acquire fire point spectral data, environmental disturbance data, and spatial positioning data to construct a raw dataset for fire point monitoring;

[0042] Among them, the fire point multispectral data includes at least thermal infrared spectral data, mid-infrared spectral data, short-wave infrared spectral data and near-infrared spectral data; the environmental disturbance data includes at least environmental temperature data, environmental humidity data, environmental air pressure data and environmental wind speed data; and the spatial positioning data is represented as the spatial positioning information of the early fire point monitoring area.

[0043] The thermal infrared spectral data is used to monitor and quantify temperature changes in the early fire detection area;

[0044] The mid-infrared spectral data is used to monitor and quantify changes in combustion gas concentration in the early fire point monitoring area;

[0045] The shortwave infrared spectral data is used to monitor and quantify changes in plant water loss in the early fire monitoring area.

[0046] The near-infrared spectral data is used to monitor and quantify changes in plant physiological activity in the early fire monitoring area;

[0047] The original fire point monitoring dataset is transformed and noise filtered to generate an initial fire point monitoring dataset. The initial fire point monitoring dataset is then standardized to obtain the final fire point monitoring dataset.

[0048] As a further aspect of the present invention, the original fire point monitoring dataset is transformed and noise filtered to generate an initial fire point monitoring dataset. The initial fire point monitoring dataset is then standardized to obtain a fire point monitoring dataset, including:

[0049] Ambient temperature evolution is performed based on thermal infrared spectral data, and environmental disturbance data is combined with compensation correction and noise filtering to generate ambient temperature monitoring data.

[0050] Combustion gas concentration is extrapolated based on mid-infrared spectral data, and compensation and correction and noise filtering are performed in combination with environmental disturbance data to generate combustion gas concentration monitoring data.

[0051] Plant water loss is extrapolated based on shortwave infrared spectral data, and environmental disturbance data is combined with compensation correction and noise filtering to generate plant water loss monitoring data.

[0052] Plant physiological activity evolution is analyzed based on near-infrared spectral data, and environmental disturbance data is combined for compensation correction and noise filtering to generate plant physiological activity monitoring data.

[0053] Among them, compensation correction means acquiring historical data, obtaining natural fluctuation residual values ​​based on historical data, and suppressing natural fluctuations based on natural fluctuation residual values; noise filtering means removing noise from interference sources based on a preset non-fire point interference source feature library.

[0054] The environmental temperature monitoring data, combustion gas concentration monitoring data, plant water loss monitoring data, and plant physiological activity monitoring data were packaged into an initial fire point monitoring dataset.

[0055] The initial fire point monitoring dataset is standardized, and the fire point monitoring dataset is spatiotemporally aligned with spatial positioning data to obtain the fire point monitoring dataset.

[0056] A multispectral early fire detection system includes:

[0057] The data acquisition module is used to collect and preprocess data from the early fire point monitoring area to generate a fire point monitoring dataset.

[0058] The data processing module is used to extract multidimensional features from the fire point monitoring dataset to obtain a multidimensional fire point feature dataset.

[0059] Fire point identification module, which is used to perform early fire point identification based on multi-dimensional collaborative fire point identification strategy and multi-dimensional fire point feature dataset, and generate early fire point identification results;

[0060] A decision execution module, which is used to execute early fire point verification decisions based on the early fire point identification results;

[0061] The feedback optimization module is used to obtain early fire point verification results and optimize the multi-dimensional collaborative fire point identification strategy based on the early fire point verification results.

[0062] Based on the above aspects, the embodiments of this application realize multi-dimensional data collection of the early fire point monitoring area, obtain the original fire point monitoring dataset, and perform data preprocessing on the original fire point monitoring dataset to obtain the fire point monitoring dataset. By collecting data from multiple directions of early fire points, the accuracy of early fire point identification is improved. Through data preprocessing operations such as compensation correction, benchmark alignment and noise filtering, noise interference caused by external interference sources is eliminated, and high-precision data input is provided for subsequent data analysis and early fire point identification.

[0063] Feature extraction is performed on the fire point monitoring dataset to obtain a multi-dimensional fire point feature dataset. The multi-dimensional fire point feature dataset includes ecological balance disturbance feature data, thermal convection motion feature data, and environmental element synergy feature data. By extracting weak signals of early fire points in a multi-dimensional manner, the accuracy of distinguishing fire points from external interference sources is improved. At the same time, multi-dimensional analysis avoids the one-sidedness and instability caused by single-dimensional analysis and provides comprehensive and reliable data support for the subsequent identification and judgment of early fire points.

[0064] Early fire point identification is performed based on a multi-dimensional collaborative fire point identification strategy and a multi-dimensional fire point feature dataset. The results of early fire point identification are obtained, and early fire point verification decisions are generated based on the results. Early fire point verification is performed based on the early fire point verification decisions, and the results of early fire point verification are obtained. The multi-dimensional collaborative fire point identification strategy is then optimized based on the results of early fire point verification. Through continuous iterative learning, the corresponding fire point identification feature weights and judgment rules in the multi-dimensional collaborative fire point identification strategy are continuously optimized, thereby ensuring high accuracy and high stability of early fire point identification. Attached Figure Description

[0065] Figure 1 This is a schematic diagram of the execution flow of a multispectral early fire point identification method provided in an embodiment of the present invention;

[0066] Figure 2 This is a schematic diagram of a multispectral early fire point identification system provided in an embodiment of the present invention;

[0067] Figure 3 This is a schematic diagram of the execution flow of step S1 in a multispectral early fire point identification method provided in an embodiment of the present invention. Detailed Implementation

[0068] The embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are used to illustrate the technical solutions of the present invention, but should not be used to limit the scope of protection of the present invention.

[0069] Example:

[0070] As attached Figure 1 , Figure 2 , Figure 3 As shown:

[0071] This invention provides a method for early fire detection based on multispectral imaging, applicable to the field of data processing technology, and includes the following steps:

[0072] Step S1: Collect multi-dimensional data from the early fire monitoring area to obtain the original fire monitoring dataset, and preprocess the original fire monitoring dataset to obtain the fire monitoring dataset.

[0073] In this embodiment, step S1 includes:

[0074] Step S11: Collect multi-dimensional data from the early fire monitoring area to obtain the original fire monitoring dataset.

[0075] Understandably, the raw dataset for fire point monitoring includes fire point multispectral data, environmental disturbance data, and spatial positioning data;

[0076] Fire point multispectral data includes at least thermal infrared spectral data, mid-infrared spectral data, short-wave infrared spectral data, and near-infrared spectral data.

[0077] Specifically, a high-resolution thermal infrared camera is used to collect thermal infrared spectral data of the early fire monitoring area, and the temperature changes in the early fire monitoring area are monitored and quantified based on the thermal infrared spectral data.

[0078] In one possible embodiment, when an early fire point appears in the early fire point monitoring area, according to Planck's law of radiation, a slight change in temperature will lead to a significant increase in the radiation energy in the corresponding wavelength band. For example, if the temperature rises from 25 degrees Celsius to 35 degrees Celsius, the radiation energy in the [8, 14] micrometer wavelength band will increase exponentially. Therefore, a small heat source can be identified by high-precision temperature measurement. For example, a small heat source of 0.03 square meters can be identified by a temperature measurement accuracy of ±0.3 degrees Celsius, thereby distinguishing the spectral characteristics of sunlight irradiation from the spectral characteristics corresponding to local combustion caused by an early fire point.

[0079] Specifically, a mid-infrared spectrometer is used to collect mid-infrared spectral data of characteristic absorption peaks of combustion gases in the early fire point monitoring area, such as collecting mid-infrared spectral data of the characteristic absorption peak of carbon dioxide at 4.2 micrometers. Based on the mid-infrared spectral data, the changes in the concentration of combustion gases in the early fire point monitoring area are monitored and quantified.

[0080] In one possible embodiment, taking carbon dioxide as an example, according to Beer-Lambert's law, the carbon dioxide released by combustion at an early ignition point will cause a decrease in the radiation intensity at the 4.2-micrometer band, and the degree of decrease is positively correlated with the concentration of carbon dioxide. By comparing the radiation intensity difference with other bands, the concentration of carbon dioxide can be inverted.

[0081] Specifically, a shortwave infrared camera is used to collect data on the plant-covered area in the early fire monitoring zone, and shortwave infrared spectral data of the plants are obtained. Based on the shortwave infrared spectral data, the changes in plant water loss in the early fire monitoring zone are monitored and quantified.

[0082] In one possible embodiment, taking hydroxyl groups in plant cell sap as an example, hydroxyl groups in plant cell sap have characteristic absorption valleys at 1.4 micrometers and 1.9 micrometers. When healthy plants have sufficient water, they will absorb light energy of the corresponding wavelength band, thereby generating the corresponding characteristic absorption valley. When plants lose water due to scorching or drought, the absorption efficiency of light energy of the corresponding wavelength band weakens, resulting in an increase in plant reflectivity. Therefore, the "pre-combustion zone" around the early fire point can be detected by the plant reflectivity image of the corresponding wavelength band. For example, if the grassland is baked and dehydrated but no open flame appears, the early fire point can be warned in advance.

[0083] Specifically, a near-infrared imager is used to collect data on the plant-covered areas in the early fire monitoring zone to obtain near-infrared spectral data. Based on the near-infrared spectral data, the changes in plant physiological activity in the early fire monitoring zone are monitored and quantified.

[0084] In one possible embodiment, taking plant chlorophyll as an example, chlorophyll absorbs light in the visible light band for photosynthesis, but in the near-infrared band, due to its spongy mesophyll structure, it has high reflectivity. When a plant is subjected to early fire-induced heat radiation, the decomposition of plant chlorophyll leads to an abnormal decrease in near-infrared reflectivity, which can indirectly reflect changes in the plant's physiological activity.

[0085] Environmental disturbance data should include at least ambient temperature data, ambient humidity data, ambient air pressure data, and ambient wind speed data.

[0086] Specifically, environmental temperature, humidity, air pressure, and wind speed data are collected in the early fire monitoring area using an environmental sensor array.

[0087] Spatial positioning data is represented as spatial positioning information of the early fire detection area.

[0088] Specifically, the spatial location information of fire point multispectral data and environmental disturbance data is recorded based on GPS and INS inertial navigation systems. The spatial location information of fire point multispectral data and environmental disturbance data is associated through time stamp synchronization to ensure that each data point can be traced back to a specific geographical location. The spatial positioning information includes latitude and longitude coordinates, altitude and time stamp.

[0089] Step S12: Perform data preprocessing on the original fire point monitoring dataset to obtain the fire point monitoring dataset.

[0090] In this embodiment, step S12 includes:

[0091] Step S12-1 involves data transformation, compensation correction, and noise filtering of the original fire point monitoring dataset.

[0092] In this embodiment, step S12-1 includes:

[0093] Steps S12-11 involve performing environmental temperature evolution analysis on the thermal infrared spectral data, and combining this data with environmental disturbance data for compensation correction and noise filtering to generate environmental temperature monitoring data.

[0094] Specifically, the collected thermal infrared spectral data is converted into brightness temperature using the radiation energy value of the thermal infrared band according to Planck's radiation law. The radiation energy values ​​of two specific atmospheric window wavelengths are selected in the thermal infrared band, such as 10.8 micrometers and 12.0 micrometers. The atmospheric water vapor content is calculated based on the radiation difference between the two bands to eliminate the artificially high temperature caused by the water vapor absorption effect.

[0095] By combining the digital elevation model with the real-time solar altitude angle, the thermal radiation attenuation coefficient of the slope-shaded area is calculated, and topographic shading compensation is performed based on the thermal radiation attenuation coefficient, such as applying 1.2 times topographic shading compensation to the north slope shaded area in the early fire monitoring area.

[0096] Based on environmental wind speed data and combined with Newton's law of cooling, the temperature difference caused by convective heat dissipation is derived, and wind speed effect compensation is performed according to the temperature difference. For example, for every 1 meter per second increase in wind speed, the resulting brightness temperature is compensated by 0.5 degrees Celsius.

[0097] Based on the brightness temperature after terrain shadow compensation and temperature compensation, a temperature field grid is constructed for the early fire point monitoring area, and the temperature field grid is output as environmental temperature monitoring data.

[0098] Step S12-12 involves extrapolating the combustion gas concentration from the mid-infrared spectral data and combining it with environmental disturbance data for compensation correction and noise filtering to generate combustion gas concentration monitoring data.

[0099] Specifically, the difference in radiation intensity between the absorption characteristic wavelength of the combustion gas and the adjacent non-absorption wavelength is compared, and the integral value of optical path concentration is calculated according to Beer-Lambert's law. For example, if the absorption characteristic wavelength of a certain combustion gas is 4.26 micrometers, the radiation intensity of the 4.26 micrometer wavelength is compared with the radiation intensity of the adjacent 4.35 micrometer non-absorption wavelength to obtain the difference in radiation intensity.

[0100] Temperature drift correction is performed based on a pre-defined library of combustion gas absorption intensity temperature coupling. For example, for every 10 degrees Celsius increase in ambient temperature, the absorption peak corresponding to the combustion gas decreases by 15%.

[0101] Aerosol optical thickness is calculated using the long-wave infrared band. Based on the aerosol optical thickness, the false absorption enhancement caused by aerosol scattering is estimated, and aerosol scattering compensation is performed based on the false absorption enhancement.

[0102] Monitor the background concentration of air upwind of the early fire point monitoring area, and subtract the baseline value of the concentration of ambient combustion gases released from non-fire points from the background concentration of air upwind.

[0103] Based on the optical path concentration integral value after noise filtering and data compensation, a combustion gas concentration distribution map of the early fire point monitoring area is constructed, and the combustion gas concentration distribution map is output as combustion gas concentration monitoring data.

[0104] Steps S12-13 involve extrapolating plant water loss from shortwave infrared spectral data and combining it with environmental disturbance data for compensation correction and noise filtering to generate plant water loss monitoring data.

[0105] In one possible embodiment, taking hydroxyl groups in plant cell sap as an example, based on... Calculate plant reflectance, where For plant reflectance, For shortwave infrared spectral data, To preset the Earth-Sun distance correction factor, Solar irradiance, Given the solar altitude angle, the normalized water sensitivity index is calculated based on (1.9 μm plant reflectance - 1.4 μm plant reflectance) / (1.9 μm reflectance + 1.4 μm reflectance). A regression model between the water sensitivity index and the actual water content is established using plant samples with known water content, thereby inverting the water sensitivity index into the amount of water loss from the plant.

[0106] In bare soil areas, a hybrid pixel decomposition method was used to remove soil spectral interference.

[0107] The plant water loss, after noise filtering and data compensation, will be output as plant water loss monitoring data.

[0108] Steps S12-14 involve analyzing the near-infrared spectral data to determine the evolution of plant physiological activity, and combining this data with environmental disturbance data for compensation correction and noise filtering to generate plant physiological activity monitoring data.

[0109] In one possible embodiment, taking plant chlorophyll as an example, high-spectral-resolution measurements are performed on the oxygen absorption band and adjacent non-absorption band corresponding to the near-infrared spectrum, the chlorophyll fluorescence intensity is extracted using the Fraunhofer line depth method, and the plant photosynthetic efficiency is estimated by the ratio of noon fluorescence efficiency to maximum fluorescence yield.

[0110] Monitor the photosynthetically active radiation intensity, establish a response curve between fluorescence signal and light intensity, and perform cloud radiation attenuation compensation based on the response curve. If the light intensity is below 500 micromoles, the attenuation compensation is activated.

[0111] By integrating observation data from multiple angles, the effective light-receiving area coefficient of plant leaves is calculated, and the light-receiving area compensation for plant photosynthetic efficiency is carried out based on the effective light-receiving area coefficient of plant leaves.

[0112] The spectral distance of the monitored plants in the early fire monitoring area was compared with that of the corresponding healthy plant samples, and external disturbances were eliminated based on the comparison results.

[0113] The plant photosynthetic efficiency, after noise filtering and data compensation, will be output as plant physiological activity monitoring data.

[0114] Furthermore, the environmental temperature monitoring data, combustion gas concentration monitoring data, plant water loss monitoring data, and plant physiological activity monitoring data are packaged and output as the initial fire point monitoring dataset.

[0115] Step S12-2: Perform data standardization on the initial fire point monitoring dataset and align the initial fire point monitoring dataset in time and space with spatial positioning data.

[0116] In this embodiment, step S12-2 includes:

[0117] Steps S12-21 involve standardizing the initial fire point monitoring dataset.

[0118] In one possible embodiment, the long-term mean and standard deviation of ambient temperature monitoring data and combustion gas concentration monitoring data in the same season and time period are calculated. The corresponding raw values ​​are converted into relative deviations in terms of standard deviations using a standard fraction conversion formula. For example, if the combustion gas concentration monitoring data is 470 ppm, it is converted to be 2.5 standard deviations higher than the baseline after standardization. For ambient temperature monitoring data, a logarithmic transformation is performed when the temperature at a single point exceeds the historical limit value to avoid extreme values ​​dominating the analysis results.

[0119] For monitoring data on plant water loss, the highest water content of healthy plants is used as the benchmark value and the lowest water content of wilted plants is used as the lower limit. The monitoring values ​​are mapped proportionally to the 0-100% range, such as 62% being converted to 0.62.

[0120] Differentiated baselines were set for plant physiological activity monitoring data based on plant type, and the plant physiological activity monitoring data were compressed to the [0,1] interval using an S-shaped function.

[0121] Steps S12-22: Perform spatiotemporal alignment on the initial fire point monitoring dataset by combining spatial positioning data.

[0122] In one possible embodiment, in the spatial dimension, the initial fire point monitoring dataset is mapped to the same reference coordinate system using a bicubic interpolation algorithm. Ground control points, such as permanent road intersection signs, are used to obtain the static installation deviation of each sensor, and an affine transformation model is applied to correct the equipment installation deviation. For example, if a sensor is offset 1.2 meters to the east, the coordinate system needs to be rotated by 0.8 degrees.

[0123] In one possible embodiment, in the time dimension, assuming that the data acquisition frequencies of ambient temperature monitoring data, combustion gas concentration monitoring data, plant water loss monitoring data, and plant physiological activity monitoring data are once per second, once every three seconds, once every two seconds, and once per second, respectively, the ambient temperature monitoring data once per second is used as the reference time axis. Linear interpolation is used to generate virtual data points per second for the combustion gas concentration monitoring data, and cubic spline interpolation is used to fit a continuous curve for the plant water loss monitoring data. The plant physiological activity monitoring data is directly synchronized to the timestamp per second.

[0124] For critical time points, such as the occurrence of thermal anomaly events, the timestamps corresponding to the combustion gas concentration monitoring data and plant water loss monitoring data are traced back and adjusted, with the start time of the temperature surge event as the zero point, to ensure that the multi-source data are strictly synchronized in the dimension of physical events and to obtain the fire point monitoring dataset.

[0125] Step S2: Extract features from the fire point monitoring dataset to obtain a multidimensional fire point feature dataset, which includes ecological balance disturbance feature data, thermal convection motion feature data, and environmental element synergistic feature data.

[0126] In this embodiment, step S2 includes:

[0127] Step S21: Analyze the time series data of the ambient temperature monitoring data contained in the fire point monitoring dataset to obtain the dynamic time series temperature change results. Based on the dynamic time series temperature change results, quantify the intensity of local thermal disturbance in the early fire point monitoring area and generate thermal convection motion characteristic data.

[0128] In one possible embodiment, for each spatial grid point in the environmental temperature monitoring data, the temperature change value compared to the previous second is calculated. For example, if the temperature rises from 25.3 degrees Celsius to 28.1 degrees Celsius, the instantaneous temperature rise is 2.8 degrees Celsius. The temperature gradient of its four adjacent grids (east, south, west, and north) is analyzed simultaneously. For example, if a grid position is found to have a temperature increase of 0.7 degrees Celsius for every 0.1 meters moved eastward and a decrease of 0.2 degrees Celsius for every 0.1 meters moved northward, a significant directional gradient is formed. When the cumulative temperature rise of a single grid point exceeds 5 degrees Celsius within five consecutive seconds or reaches more than three times the historical standard deviation of the area, and the spatial gradient direction is consistent, such as if it continues to increase eastward, the spatial grid point is marked as a thermal anomaly point. For a group of consecutive anomaly points in space, the grid area corresponding to the anomaly point group is defined as the thermal anomaly core area and its outline boundary is marked. For example, if all 25 spatial grid points in a 5×5 grid area are thermal anomaly points, the corresponding grid area is defined as the thermal anomaly core area and its outline boundary is marked.

[0129] Assuming a 0.5m × 0.5m analysis unit, calculate the curl of the velocity vector of all grid points within the unit. If the average curl of the velocity vector is measured to be a counterclockwise rotation at a rotational angular velocity of 0.3 radians / second, calculate the eigenvalues ​​of the velocity gradient matrix. When the ratio of rotation intensity to deformation rate exceeds the critical threshold, it is determined that a thermal vortex structure has been formed.

[0130] For the detected vortex core, its positional shift in consecutive frames is tracked. For example, if it moves from X35.6-Y122.1 to X37.2-Y123.9 within ten seconds, the vortex expansion rate is 1.08 meters per minute in the northeast direction. Its rotational angular velocity change is also statistically analyzed, such as accelerating from an initial 0.3 radians / second to 0.5 radians / second. The ratio of the cumulative number of newly added vortices within every 30-second window to the initial base number is used as the growth rate. For example, if the initial number of vortices is 10 and the number now increases to 25, the corresponding growth rate is 150%.

[0131] The number of vortex entities, the dynamic characteristics of vortex groups, the vortex motion trend, and the vortex motion direction are output as thermal convection motion characteristic data. For example, a certain thermal convection motion characteristic data can be represented as [number of vortex entities: 8, vortex group dynamic characteristics: {average rotational angular velocity is 0.4 radians / second, maximum rotational intensity is 0.6 radians / second}, vortex group motion trend: {average expansion rate is 0.9 meters per minute, maximum single vortex expansion rate is 1.8 meters per minute}, vortex motion direction: {70% of vortices move in the northeast direction, 20% of vortices move in the southeast direction, and 10% of vortices move in other directions}].

[0132] Step S22: Based on the combustion gas concentration monitoring data, plant water loss monitoring data, and plant physiological activity monitoring data contained in the fire point monitoring dataset, construct an ecological balance disturbance correlation map, and generate ecological balance disturbance characteristic data by analyzing the temporal interaction correlation of the ecological balance disturbance correlation map.

[0133] In one possible embodiment, a sudden increase in the monitoring data value of combustion gas concentration is regarded as an environmental pollution pressure source node, a decrease in the monitoring data value of plant water loss is regarded as a physiologically related response node, and a decrease in the monitoring data value of plant physiological activity is regarded as an ecological function damage node.

[0134] In each minute-by-minute time-series slice, the transmission delay from environmental pollution pressure source nodes to physiologically related response nodes is calculated using cross-correlation function analysis. For example, plant water content loss occurs synchronously 3 seconds after a combustion gas concentration exceedance event. The coupling strength between physiologically related response nodes and ecological function damage nodes is quantified using the Jaccard similarity index. For example, if 85% of the water loss area experiences synchronous decline in plant physiological activity, the corresponding coupling strength is 0.85. The degree of loss during the transmission process from environmental pollution pressure source nodes to physiologically related response nodes is obtained, and the response amplitude attenuation rate is obtained based on the degree of loss. For example, if the gas concentration exceeds the standard by 10 ppm, the expected water decrease should be 3%, but the actual decrease is only 1.8%, then the response amplitude attenuation rate is 0.6. The product of the response amplitude attenuation rate and the coupling strength is used as the single-link transmission efficiency to quantify the signal transmission effectiveness from environmental pollution pressure source nodes to terminal ecological function damage nodes.

[0135] The coverage area of ​​the conduction chain corresponding to the environmental pollution pressure source node and the physiologically related response node is defined as the gas moisture conduction area, and the coverage area of ​​the conduction chain corresponding to the physiologically related response node and the ecological function damage node is defined as the moisture physiological damage conduction area. The spatial overlap between the gas moisture conduction area and the moisture physiological damage conduction area is calculated, and the overlap rate between links is obtained based on the spatial overlap. If the gas moisture conduction area and the moisture physiological damage conduction area occupy 50 grid areas at the same time, and 45 grid areas in the moisture physiological damage conduction area spatially overlap with the gas moisture conduction area, then the corresponding overlap rate between links is 90%.

[0136] When a complete conduction chain with a single-link conduction efficiency greater than 0.7 and an inter-link overlap rate greater than 90% is detected in five consecutive time slices, the completeness of the conduction network, the intensity of node imbalance, and the system lag time are generated and output as ecological balance disturbance feature data. For example, the single-link conduction efficiency corresponding to "excessive gas concentration leads to water loss in plants, resulting in a decline in plant physiological activity" is 0.75, and the inter-link overlap rate is 95%, which is determined to be a complete conduction chain, and the corresponding ecological balance disturbance feature data are generated.

[0137] Understandably, the completeness of the conduction network is represented by the proportion of effective conduction paths to theoretical paths, the node imbalance intensity is represented by the average standard deviation of environmental pollution pressure source nodes, physiologically related response nodes, and ecological function damage nodes from the healthy baseline, and the system lag time is represented by the average time from the start of a sudden increase in gas concentration until the plant's physiological activity decreases.

[0138] Step S23: Based on the environmental causal coupling mechanism, analyze the environmental disturbance data and plant water loss monitoring data contained in the fire point monitoring dataset to generate environmental element synergistic feature data.

[0139] Specifically, by reconstructing and integrating time-series data of environmental disturbance data and plant water loss monitoring data in phase space, and generating corresponding phase space motion trajectories, the chaotic feature data corresponding to the abrupt changes in phase space motion trajectory caused by early fire points are obtained.

[0140] The causal strength between environmental disturbance data and plant water loss monitoring data is calculated based on the convergent cross-mapping algorithm. A causal feature network is constructed based on the causal transmission path and corresponding causal strength between the environmental disturbance data and plant water loss monitoring data. Causal feature data is obtained based on the causal feature network.

[0141] The fluctuation phase of environmental disturbance data and plant water loss monitoring data was extracted by Hilbert transform, and phase-coordinated feature data was obtained based on the fluctuation phase.

[0142] Chaotic feature data, causal feature data, and phase-coordinated feature data are encapsulated and output as environmental element coordinated feature data.

[0143] Step S24: Construct a multidimensional fire point feature dataset based on thermal convection motion feature data, ecological balance disturbance feature data, and environmental element collaborative feature data.

[0144] Specifically, thermal convection motion characteristic data, ecological balance disturbance characteristic data, and environmental element synergistic characteristic data are encapsulated to generate a multidimensional fire point characteristic dataset.

[0145] Step S3: Based on the multi-dimensional collaborative fire point recognition strategy and the multi-dimensional fire point feature dataset, perform early fire point recognition and obtain the early fire point recognition results.

[0146] In this embodiment, step S3 includes:

[0147] Step S31: Locate the thermodynamic anomaly region by analyzing thermal convection motion characteristic data.

[0148] Specifically, the thermal convection motion characteristic data is decomposed into three physical quantity fields: velocity field, acceleration field, and vorticity field. The expansion mode feature is extracted through a spatiotemporal convolutional network to obtain expansion mode feature data. Based on the expansion mode feature data, the fire point thermal plume index is generated. When the fire point thermal plume index is greater than the preset thermodynamic pressure threshold, it is determined that a thermodynamic anomaly has occurred in the early fire point monitoring area, and the spatial location of the corresponding thermodynamic anomaly in the monitoring area is obtained.

[0149] Among them, the velocity field is represented by the displacement vector per second of the core point of the vortex group corresponding to the local thermal disturbance, the acceleration field is represented by the acceleration per second calculated based on the velocity field difference, which is used to identify abnormal acceleration zones, and the vorticity field is represented by the product of the vortex rotation intensity and the vortex area, which is used to quantify the rotational kinetic energy density of the vortex.

[0150] In one possible embodiment, assuming that in an early fire detection zone, three vortex core points A1, B2, and C3 are detected through thermal convection motion characteristic analysis, during a three-second observation period, in the first second, A1 moves northeast at a speed of 0.8 meters per second, B2 moves east at a speed of 0.7 meters per second, and C3 moves southeast at a speed of 0.5 meters per second; in the second second, A1's speed suddenly increases to 1.3 meters per second, with a corresponding acceleration of 0.5 meters per second squared; B2 maintains a constant speed, with a corresponding acceleration of 0 meters per second squared; C3's speed suddenly increases to 1.3 meters per second, and its direction deflects to the northeast, with a calculated corresponding acceleration of 0.6 meters per second squared. Since the accelerations of A1 and C3 are greater than a preset threshold, A1 and C3 are determined to be abnormal. Acceleration zone; the rotation intensity of A1 is obtained as 0.4 radians / second, and the coverage area is 2 square meters, so the corresponding vorticity value is 0.4 × 2 = 0.8 joules / square meter. The rotation intensity of C3 is obtained as 0.5 radians / second, and the coverage area is 3 square meters, so the corresponding vorticity value is 1.5 joules / square meter. Therefore, the average vorticity value of A1 and C3 is (0.8 + 1.5) / 2 = 1.15 joules / square meter. By analyzing the three physical fields of A1 and C3 through spatiotemporal convolution kernel, a ring shock wave structure with a closed ring distribution of acceleration vector is detected at the C3 position. The corresponding shock wave intensity is obtained as 0.9, and the expansion correlation is 0.85, where the expansion correlation indicates the consistency of the acceleration direction of A1 and C3. According to "Fire point thermal plume index = shock wave intensity * expansion coherence * The fire point thermal plume index is calculated using the function form "ln(1+average vorticity value)" and is approximately 0.585 (0.9*0.85*ln(1+1.15)). Since 0.585 is greater than the preset fire point thermal plume index threshold, it is determined that a thermodynamic anomaly has occurred, and the spatial positions corresponding to A1 and C3 are output.

[0151] Step S32: Perform fire point inference for the thermodynamic anomaly area.

[0152] Specifically, a multi-dimensional collaborative fire point correlation map is constructed based on the ecological balance disturbance characteristic data and environmental element synergistic characteristic data corresponding to the thermodynamic anomaly area. Fire point inference is then performed based on the multi-dimensional collaborative fire point correlation map to obtain the fire point inference results.

[0153] In one possible embodiment, the above A1 and C3 cases are used as a reference. Environmental element collaborative analysis is performed on region A1. In the chaotic feature extraction stage, the environmental temperature fluctuation and vegetation water loss time series are mapped to a three-dimensional state trajectory of temperature standard deviation, water loss rate and loss acceleration through phase space reconstruction. The original circular trajectory is captured to collapse and bifurcate towards the low water zone within 32 seconds. The fractal dimension value of 1.3 and the Lyapunov exponent of 0.12 are quantitatively output. The Lyapunov exponent of 0.12 is higher than the preset instability threshold of 0.05.

[0154] In the causal feature construction stage, the convergent cross-mapping algorithm was used to verify the directional driving effect of temperature fluctuation on water loss. The calculated conduction strength between temperature and plant water reached 0.88, with a response lag of 3 seconds, forming a single dominant causal link.

[0155] In the phase coordination calculation stage, the phase difference between temperature and moisture fluctuations was analyzed based on Hilbert transform, and the phase synchronization index was measured to be 0.93, confirming cross-element coordinated disturbance. The above ternary features were encapsulated into environmental element coordinated feature data {fractal dimension value 1.3, Lyapunov index 0.12, temperature and moisture conduction intensity 0.88, phase synchronization index 0.93}. The corresponding ecological balance disturbance feature data were obtained, in which the conduction network completeness was 95% and the node imbalance intensity was 2.8σ. Based on the environmental element coordinated feature data and the ecological balance disturbance feature data, fire point inference was performed. Since the fractal dimension value of 1.3 is lower than the preset critical value of 1.5, the temperature and moisture conduction intensity is 0.88, and strong causal conduction is shown between temperature and moisture, and the phase synchronization index of 0.93 represents the disturbance coordination, the final inference is that a fire point ecological collapse has occurred, and the probability of the fire point occurrence is 96%. The core coordinates of A1 (3.2E, 2.5N) and the warning of the extended hot zone are output.

[0156] Fire point inference was performed based on the ecological balance disturbance characteristic data and environmental element synergy characteristic data corresponding to region C3. The fractal dimension value of region C3 is 1.8, which indicates a stable state. The temperature and moisture conduction intensity is 0.30, indicating no dominant path. The phase synchronization index is 0.40, showing randomness. Combined with the low conduction network completeness value of only 40%, the fire point threat was ruled out, and the inference was that it was caused by natural drought.

[0157] Step S33: Generate early fire point identification results based on the fire point thermal plume index and fire point inference results.

[0158] Obtain the spatial location of the inference association region associated with the fire point inference result, and perform spatial overlap analysis on the spatial location of the thermodynamic anomaly and the inference association spatial location to obtain the spatial overlap.

[0159] When the spatial overlap is not lower than the preset overlap threshold, it is determined that an early fire point has occurred. The risk level is generated based on the fire point thermal plume index and the fire point inference result, and the spatial location and risk level corresponding to the early fire point are output.

[0160] When the spatial overlap is lower than the preset overlap threshold, it is determined that no early fire point has occurred.

[0161] In one possible embodiment, following the examples A1 and C3 above, a gridded calculation method is used to divide the target area into grids with a resolution of 0.1 meters. For example, a 10-meter × 10-meter early fire monitoring area is planned to be divided into 10,000 grids. The proportion of grids that simultaneously fall into the thermal anomaly area corresponding to the spatial location of the thermal anomaly and the inference associated area corresponding to the spatial location of the inference association is counted. For example, if the thermal anomaly area and the inference associated area share 170 grids and the total number of covered grids is 200, the corresponding spatial overlap is 85%. Assuming that the preset overlap threshold is 80%, at this time 85% exceeds 80%, it is determined that an early fire has occurred in the area. The risk level is generated based on the fire plume index and the probability of fire occurrence in the fire inference results. Since the fire plume index of A1 is greater than 0.4 and the probability of fire occurrence exceeds 95%, it is determined to be the highest risk level. The core coordinates of A1 are output, and the high-risk area of ​​northeastward diffusion is marked.

[0162] Step S4: Perform early fire point verification decision based on the early fire point identification result and obtain the early fire point verification result.

[0163] Specifically, the early fire point identification results are analyzed to obtain the risk level of the early fire point;

[0164] The risk levels include high risk, medium risk, and low risk.

[0165] Decisions for early fire detection are made based on risk level:

[0166] For early fire points in high-risk and medium-risk areas, multi-source collaborative verification of early fire points is carried out, such as increasing the acquisition frequency of corresponding data acquisition equipment and conducting fire tracking and analysis. At the same time, drones equipped with higher-precision data acquisition equipment are dispatched to assist in analysis and verification, and fire point warning information is generated to the terminals of patrol personnel around the corresponding early fire point area, so that the patrol personnel can conduct fire point investigation in the early fire point area.

[0167] For low-risk early fire point areas, manual verification is carried out. For example, the information corresponding to low-risk early fire point areas is sent to relevant departments, which then formulate inspection tasks and regularly assign inspection personnel to carry out inspection tasks. When carrying out inspection tasks, inspection personnel can conduct early fire point investigation in low-risk early fire point areas.

[0168] Early fire point verification results are obtained through early fire point verification decisions, and fire point elimination strategies are executed based on these results.

[0169] Understandably, the fire suppression strategy means that for high-risk early fire areas identified as successfully identified in the early fire verification results, fire spread suppression is implemented, and fires are eliminated based on the effectiveness of the fire spread suppression. For example, fire suppression can be carried out initially using fire suppression devices mounted on drones, and fire information can be sent to relevant departments to assist them in assigning professional personnel and corresponding equipment to extinguish the fires. For medium-risk early fire areas identified as successfully identified as successfully identified, fire information can be sent to relevant departments to assist them in assigning professional personnel and corresponding equipment to extinguish the fires.

[0170] In one possible embodiment, following the examples of A1 and C3 above, since area A1 is determined to be a high-risk area, early multi-source collaborative verification of fire points is required. The following verification strategy can be implemented: dispatching drones to detect the concentration of characteristic gases of fire points and simultaneously verifying the persistence and diffusion direction of heat sources.

[0171] The fire point type was verified as a real fire with smoldering characteristics, and the fire point spread trend was northeastward. Based on the verification results of the fire point type and the fire point spread trend, the fire spread was suppressed by positioning the coordinates in area A1. The fire suppression effect was found to be 90% initial open flame extinguishing rate. The residual smoldering needs to be continuously monitored. The fire point was determined to be successfully identified.

[0172] Step S5: Optimize the multi-dimensional collaborative fire point identification strategy based on the early fire point verification results.

[0173] In this embodiment, step S5 includes:

[0174] Step S51: Strengthen the confidence of the rules contained in the multi-dimensional collaborative fire point identification strategy corresponding to the successful early fire point identification in the early fire point verification results.

[0175] In one possible embodiment, a consistency score for multi-source verification data is generated based on the early fire point verification results. For example, the spatiotemporal alignment of combustion gas verification, temperature field verification, and fire spread emergency suppression response can be used as the scoring criteria. The higher the spatiotemporal alignment, the greater the utility of the corresponding rule in the multi-dimensional collaborative fire point identification strategy. The rule is then marked as a high-confidence rule.

[0176] Step S52: Perform interference source diagnosis on cases where early fire point identification fails in the early fire point verification results, obtain interference source diagnosis results, and expand the interference source diagnosis results to the preset non-fire point interference source feature library.

[0177] In one possible embodiment, the original multispectral data corresponding to the failed cases are matched with a preset non-fire point interference source feature library to identify potential interference sources, such as the spectral single-peak shape of industrial heat sources and the potassium characteristic absorption valley of agricultural burning; the spatiotemporal fluctuation characteristics of the multidimensional fire point feature dataset corresponding to the failed cases are analyzed to decouple the difference between real fire points and interference events; features of new interference sources are extracted, interference source feature templates are generated and implanted into the preset non-fire point interference source feature library; the initial weight of the newly implanted interference source feature template in the preset non-fire point interference source feature library is 0.7, and the corresponding weight of the interference source feature template is increased by 0.05 for each correct interception of a false alarm, with an upper limit of 0.95; inefficient interference source feature templates that have not been hit for 6 consecutive months are removed.

[0178] This invention provides a multispectral early fire detection system, applicable to the field of data processing technology, including:

[0179] The data acquisition module is used to collect and preprocess data from the early fire point monitoring area to generate a fire point monitoring dataset.

[0180] The data processing module is used to extract multidimensional features from the fire point monitoring dataset to obtain a multidimensional fire point feature dataset.

[0181] Fire point identification module, which is used to perform early fire point identification based on multi-dimensional collaborative fire point identification strategy and multi-dimensional fire point feature dataset, and generate early fire point identification results;

[0182] A decision execution module, which is used to execute early fire point verification decisions based on the early fire point identification results;

[0183] The feedback optimization module is used to obtain early fire point verification results and optimize the multi-dimensional collaborative fire point identification strategy based on the early fire point verification results.

[0184] The specific usage and function of this embodiment are as follows:

[0185] First, multi-dimensional data collection is carried out in the early fire point monitoring area to obtain the original fire point monitoring dataset. Then, the original fire point monitoring dataset is preprocessed to obtain the fire point monitoring dataset. By collecting data from multiple angles of early fire points, the one-sidedness caused by analyzing a single data point is avoided. Through data preprocessing operations such as compensation correction, benchmark alignment, and noise filtering, noise interference from external interference sources is eliminated. At the same time, high-precision and reliable data is provided for subsequent data analysis and early fire point identification.

[0186] Next, feature extraction is performed on the fire point monitoring dataset to obtain a multi-dimensional fire point feature dataset. The multi-dimensional fire point feature dataset includes ecological balance disturbance feature data, thermal convection motion feature data, and environmental element synergy feature data. By extracting features in multiple dimensions, the weak signals of early fire points are captured, thereby accurately distinguishing fire points from external interference sources. This improves the accuracy and timeliness of early fire point identification and provides a comprehensive and reliable basis for subsequent early fire point identification and judgment.

[0187] Finally, early fire point identification is performed based on a multi-dimensional collaborative fire point identification strategy and a multi-dimensional fire point feature dataset to obtain early fire point identification results. Early fire point verification decisions are generated based on the early fire point identification results, and early fire point verification is performed based on the early fire point verification decisions to obtain early fire point verification results. The multi-dimensional collaborative fire point identification strategy is then optimized based on the early fire point verification results. Through continuous iterative learning, the corresponding fire point identification feature weights and judgment rules are continuously optimized to maintain high accuracy and high stability of early fire point identification.

[0188] Furthermore, embodiments of the present invention also provide an electronic device, the electronic device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the methods described above.

[0189] The following is a detailed introduction to the various components of the electronic device:

[0190] In this context, the processor is the control center of the electronic device. It can be a single processor or a collective term for multiple processing elements. For example, a processor can be one or more central processing units (CPUs), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention, such as one or more digital signal processors (DSPs) or one or more field-programmable gate arrays (FPGAs).

[0191] The processor can perform various functions of an electronic device by running or executing software programs stored in memory and by calling data stored in memory.

[0192] The memory is used to store the software program that executes the solution of the present invention, and the execution is controlled by the processor. For specific implementation methods, please refer to the above method embodiments, which will not be repeated here.

[0193] The memory can be a real-only memory (ROM) or other type of static storage device capable of storing static information and instructions, a random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only (CD-ROM), or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media, or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto. The memory can be integrated with the processor or exist independently and coupled to the processor through an interface circuit of an electronic device; this embodiment of the invention does not specifically limit this.

[0194] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via limited means (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0195] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0196] It should be understood that, in the embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0197] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for early fire point identification based on multispectral imaging, characterized in that, The method includes: Multi-dimensional data collection was conducted on the early fire point monitoring area to obtain the raw fire point monitoring dataset, and the raw fire point monitoring dataset was preprocessed to obtain the fire point monitoring dataset. Feature extraction is performed on the fire point monitoring dataset to obtain a multidimensional fire point feature dataset, which includes ecological balance disturbance feature data, thermal convection motion feature data, and environmental element synergy feature data. The feature extraction of the fire point monitoring dataset refers to the analysis of the time series data of the environmental temperature monitoring data contained in the fire point monitoring dataset, obtaining the dynamic time series temperature change results, quantifying the local thermal disturbance intensity in the early fire point monitoring area based on the dynamic time series temperature change results, and generating thermal convection motion characteristic data. Based on the combustion gas concentration monitoring data, plant water loss monitoring data, and plant physiological activity monitoring data contained in the fire point monitoring dataset, an ecological balance disturbance correlation map is constructed, and ecological balance disturbance characteristic data is generated by analyzing the temporal interaction correlation of the ecological balance disturbance correlation map. Based on the environmental causal coupling mechanism, the environmental disturbance data and plant water loss monitoring data contained in the fire point monitoring dataset are analyzed to generate environmental element synergistic feature data. A multidimensional fire point feature dataset was constructed based on thermal convection motion characteristic data, ecological balance disturbance characteristic data, and environmental element synergistic characteristic data. Early fire point identification is performed based on a multi-dimensional collaborative fire point identification strategy and a multi-dimensional fire point feature dataset to obtain early fire point identification results. Based on the early fire point identification results, perform early fire point verification decisions and obtain the early fire point verification results; The multi-dimensional collaborative fire point identification strategy was optimized based on the early fire point verification results.

2. The method for early fire point identification based on multispectral imaging according to claim 1, characterized in that, Based on the environmental causal coupling mechanism, the environmental disturbance data and plant water loss monitoring data contained in the fire point monitoring dataset are analyzed to generate synergistic characteristic data of environmental elements, including: By reconstructing and integrating time-series data of environmental disturbance data and plant water loss monitoring data in phase space, and generating corresponding phase space motion trajectories, chaotic feature data corresponding to the sudden changes in phase space motion trajectory caused by early fire points are obtained. The causal strength between environmental disturbance data and plant water loss monitoring data is calculated based on the convergent cross-mapping algorithm. A causal feature network is constructed based on the causal transmission path and corresponding causal strength between the environmental disturbance data and plant water loss monitoring data. Causal feature data is obtained based on the causal feature network. The fluctuation phase of environmental disturbance data and plant water loss monitoring data was extracted by Hilbert transform, and phase-coordinated feature data was obtained based on the fluctuation phase. Chaotic feature data, causal feature data, and phase-coordinated feature data are encapsulated and output as environmental element coordinated feature data.

3. The method for early fire point identification based on multispectral imaging according to claim 1, characterized in that, Early fire point identification is performed based on a multi-dimensional collaborative fire point identification strategy and a multi-dimensional fire point feature dataset. The results of early fire point identification are obtained, including: The thermal convection motion characteristic data is decomposed into velocity field, acceleration field and vortex field, and the expansion mode feature is extracted by spatiotemporal convolutional network to obtain expansion mode feature data. Based on the expansion mode feature data, the fire point thermal plume index is generated. When the fire point thermal plume index is greater than the preset thermodynamic pressure threshold, it is determined that a thermodynamic anomaly has occurred in the early fire point monitoring area, and the spatial location of the thermodynamic anomaly corresponding to the early fire point monitoring area is obtained. A multidimensional collaborative fire point correlation map is constructed based on the ecological balance disturbance characteristic data and environmental element synergistic characteristic data corresponding to the thermodynamic anomaly area. Fire point inference is performed based on the multidimensional collaborative fire point correlation map to obtain the fire point inference results. Early fire point identification results are generated based on the fire point thermal plume index and fire point inference results.

4. The method for early fire point identification based on multispectral imaging according to claim 3, characterized in that, Early fire point identification results are generated based on the fire point thermal plume index and fire point inference results, including: Obtain the spatial location of the inference association region associated with the fire point inference result, and perform spatial overlap analysis on the spatial location of the thermodynamic anomaly and the inference association spatial location to obtain the spatial overlap. When the spatial overlap is not lower than the preset overlap threshold, it is determined that an early fire point has occurred. The risk level is generated based on the fire point thermal plume index and the fire point inference result, and the spatial location and risk level corresponding to the early fire point are output. When the spatial overlap is lower than the preset overlap threshold, it is determined that no early fire point has occurred.

5. The method for early fire point identification based on multispectral imaging according to claim 1, characterized in that, Based on the early fire point identification results, perform early fire point verification decisions and obtain the early fire point verification results, including: Analyze the early fire point identification results to obtain the risk level of the early fire point; The risk levels include high risk, medium risk, and low risk. Decisions for early fire detection are made based on risk level: For early fire point areas at high and medium risk levels, multi-source collaborative verification of early fire points is conducted. For early fire zones with low risk levels, manual verification is conducted. Early fire point verification results are obtained through early fire point verification decisions, and fire point elimination strategies are executed based on these results.

6. The method for early fire point identification based on multispectral imaging according to claim 1, characterized in that, The multi-dimensional collaborative fire detection strategy was optimized based on early fire detection results, including: The confidence level of the rules contained in the multi-dimensional collaborative fire point identification strategy corresponding to the successful early fire point identification in the early fire point verification results is enhanced. For cases where early fire point identification fails in the early fire point verification results, interference source diagnosis is performed to obtain the interference source diagnosis results, and the interference source diagnosis results are expanded to the preset non-fire point interference source feature library.

7. The method for early fire point identification based on multispectral imaging according to claim 1, characterized in that, Multi-dimensional data collection was conducted in the early fire detection area to obtain the raw fire detection dataset. This raw dataset was then preprocessed to obtain the final fire detection dataset, which includes: Acquire fire point spectral data, environmental disturbance data, and spatial positioning data to construct a raw dataset for fire point monitoring; Among them, the fire point multispectral data includes at least thermal infrared spectral data, mid-infrared spectral data, short-wave infrared spectral data and near-infrared spectral data; the environmental disturbance data includes at least environmental temperature data, environmental humidity data, environmental air pressure data and environmental wind speed data; and the spatial positioning data is represented as the spatial positioning information of the early fire point monitoring area. The thermal infrared spectral data is used to monitor and quantify temperature changes in the early fire detection area; The mid-infrared spectral data is used to monitor and quantify changes in combustion gas concentration in the early fire point monitoring area; The shortwave infrared spectral data is used to monitor and quantify changes in plant water loss in the early fire monitoring area. The near-infrared spectral data is used to monitor and quantify changes in plant physiological activity in the early fire monitoring area; The original fire point monitoring dataset is transformed and noise filtered to generate an initial fire point monitoring dataset. The initial fire point monitoring dataset is then standardized to obtain the final fire point monitoring dataset.

8. The method for early fire point identification based on multispectral imaging according to claim 7, characterized in that, The original fire point monitoring dataset is transformed and noise filtered to generate an initial fire point monitoring dataset. This initial fire point monitoring dataset is then standardized to obtain the final fire point monitoring dataset, which includes: Ambient temperature evolution is performed based on thermal infrared spectral data, and environmental disturbance data is combined with compensation correction and noise filtering to generate ambient temperature monitoring data. Combustion gas concentration is extrapolated based on mid-infrared spectral data, and compensation and correction and noise filtering are performed in combination with environmental disturbance data to generate combustion gas concentration monitoring data. Plant water loss is extrapolated based on shortwave infrared spectral data, and environmental disturbance data is combined with compensation correction and noise filtering to generate plant water loss monitoring data. Plant physiological activity evolution is analyzed based on near-infrared spectral data, and environmental disturbance data is combined for compensation correction and noise filtering to generate plant physiological activity monitoring data. Among them, compensation correction means acquiring historical data, obtaining natural fluctuation residual values ​​based on historical data, and suppressing natural fluctuations based on natural fluctuation residual values; noise filtering means removing noise from interference sources based on a preset non-fire point interference source feature library. The environmental temperature monitoring data, combustion gas concentration monitoring data, plant water loss monitoring data, and plant physiological activity monitoring data were packaged into an initial fire point monitoring dataset. The initial fire point monitoring dataset is standardized, and the fire point monitoring dataset is spatiotemporally aligned with spatial positioning data to obtain the fire point monitoring dataset.

9. A multispectral early fire detection system for implementing the method described in any one of claims 1 to 8, characterized in that, The system includes: The data acquisition module is used to collect and preprocess data from the early fire point monitoring area to generate a fire point monitoring dataset. The data processing module is used to extract multidimensional features from the fire point monitoring dataset to obtain a multidimensional fire point feature dataset. Fire point identification module, which is used to perform early fire point identification based on multi-dimensional collaborative fire point identification strategy and multi-dimensional fire point feature dataset, and generate early fire point identification results; A decision execution module, which is used to execute early fire point verification decisions based on the early fire point identification results; The feedback optimization module is used to obtain early fire point verification results and optimize the multi-dimensional collaborative fire point identification strategy based on the early fire point verification results.

Images