Fire point determination method and apparatus, electronic device, and computer readable storage medium

Abstract

The application discloses a fire point determination method and device, electronic equipment and a computer readable storage medium. The method comprises the following steps: obtaining a target image of a target location and a historical fire point multidimensional probability image corresponding to the target location; obtaining an initial fire point probability image according to the target image, wherein the initial fire point probability image is an image of an initial probability of a fire point appearing at a corresponding position of a pixel point in the target image; and obtaining a target fire point probability image according to the target image, the historical fire point multidimensional probability image and the initial fire point probability image, wherein the target fire point probability image is an image of a target probability of a fire point appearing at a corresponding position of a pixel point in the target image. The application solves the technical problem that the determined fire point probability is inaccurate in the related art when determining the fire point probability of a certain position.

Description

Technical Field

[0001] This invention relates to the field of detection, and more specifically, to a method, apparatus, electronic device, and computer-readable storage medium for determining a fire point. Background Technology

[0002] Fire detection utilizes remote sensing technology to observe large areas and employs intelligent data analysis to determine the location of fires. Fire detection is crucial for power grid safety and forest protection. However, in related technologies, determining the probability of a fire at a specific location can be inaccurate.

[0003] There is currently no effective solution to the above problems. Summary of the Invention

[0004] This invention provides a method, apparatus, electronic device, and computer-readable storage medium for determining fire points, in order to at least solve the technical problem in the related art where the determined fire point probability at a certain location is inaccurate.

[0005] According to one aspect of the present invention, a method for determining fire points is provided, comprising: acquiring a target image of a target location and a historical fire point multi-dimensional probability image corresponding to the target location, wherein the historical fire point multi-dimensional probability image is an image displaying the historical probability of fire points appearing at corresponding positions of pixels in the image in multiple dimensions; obtaining an initial fire point probability image based on the target image, wherein the initial fire point probability image is an image displaying the initial probability of fire points appearing at corresponding positions of pixels in the target image; and obtaining a target fire point probability image based on the target image, the historical fire point multi-dimensional probability image, and the initial fire point probability image, wherein the target fire point probability image is an image displaying the target probability of fire points appearing at corresponding positions of pixels in the target image.

[0006] Optionally, obtaining the target fire point probability image based on the target image, the historical fire point multi-dimensional probability image, and the initial fire point probability image includes: inputting the target image, the historical fire point multi-dimensional probability image, and the initial fire point probability image into a target fire point detection model to obtain the target fire point probability image. The target fire point detection model is obtained by training an initial fire point detection model based on sample pair data. The sample pair data includes positive sample images, historical positive sample fire point multi-dimensional probability images, initial positive sample fire point probability images, target positive sample fire point probability images, and negative sample images, historical negative sample fire point multi-dimensional probability images, initial negative sample fire point probability images, and target negative sample fire point probability images. The target positive sample fire point probability image is an image where the probability of a fire point appearing at the corresponding position of a pixel in the image is greater than a first threshold. The target negative sample fire point probability image is an image where the probability of a fire point appearing at the corresponding position of a pixel in the image is less than a second threshold, where the first threshold is greater than the second threshold.

[0007] Optionally, before inputting the target image, the historical fire point multi-dimensional probability image, and the initial fire point probability image into the target fire point detection model to obtain the target fire point probability image, the method further includes: obtaining the initial fire point detection model and constructing a loss function for model training, wherein the loss function includes a cross-entropy loss function; and training the initial fire point detection model using the sample pair data based on the loss function to obtain the target fire point detection model.

[0008] Optionally, obtaining the target fire point probability image based on the target image, the historical fire point multi-dimensional probability image, and the initial fire point probability image includes: determining the probability clustering result of fire points appearing at corresponding positions of pixels in the target image based on the historical fire point multi-dimensional probability image and the initial fire point probability image; and obtaining the target fire point probability image based on the probability clustering result of fire points appearing at corresponding positions of pixels in the target image.

[0009] Optionally, before acquiring the target image of the target location, the method further includes: acquiring an initial image of the target location; registering the initial image according to predetermined parameter information to obtain the target image, wherein the predetermined parameter information includes a predetermined resolution and spatial corresponding position information of each vertex of the image.

[0010] Optionally, obtaining a multi-dimensional probability image of historical fire points corresponding to the target location includes: obtaining historical fire point data of the target location; and determining the multi-dimensional probability image of historical fire points corresponding to the target location based on the historical fire point data, wherein the number of dimensions of the multi-dimensional image is determined according to the number of types of historical fire point data.

[0011] Optionally, obtaining the initial fire point probability image based on the target image includes: determining the initial fire point probability image based on the target image and the target algorithm, wherein the target algorithm includes the Constrained Energy Minimization (CEM) algorithm.

[0012] According to one aspect of the present invention, a fire point determination device is provided, comprising: a first acquisition module, configured to acquire a target image of a target location and a historical fire point multi-dimensional probability image corresponding to the target location, wherein the historical fire point multi-dimensional probability image is an image displaying the historical probability of fire points appearing at corresponding positions of pixels in the image in multiple dimensions; a second acquisition module, configured to obtain an initial fire point probability image based on the target image, wherein the initial fire point probability image is an image displaying the initial probability of fire points appearing at corresponding positions of pixels in the target image; and a third acquisition module, configured to obtain a target fire point probability image based on the target image, the historical fire point multi-dimensional probability image, and the initial fire point probability image, wherein the target fire point probability image is an image displaying the target probability of fire points appearing at corresponding positions of pixels in the target image.

[0013] According to one aspect of the present invention, an electronic device is provided, comprising: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to execute the instructions to implement the fire point determination method described in any of the preceding claims.

[0014] According to one aspect of the present invention, a computer-readable storage medium is provided, wherein when the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to perform the fire point determination method described in any of the preceding claims.

[0015] In this embodiment of the invention, a target image of the target location and a historical multi-dimensional probability image of fire points corresponding to the target location are first acquired. The historical multi-dimensional probability image of fire points displays the historical probability of fire points appearing at corresponding pixel positions in multiple dimensions within the image. Then, based on the target image, an initial fire point probability image is obtained. This initial fire point probability image displays the initial probability of fire points appearing at corresponding pixel positions within the target image. Finally, based on the target image, the historical multi-dimensional probability image of fire points, and the initial fire point probability image, a target fire point probability image is obtained. This target fire point probability image displays the target probability of fire points appearing at corresponding pixel positions within the target image. Since the acquired target image is an image of the target location, it accurately reflects the current situation at the target location. Therefore, the initial fire point probability image obtained based on the target image can accurately display the initial probability of fire points appearing at corresponding pixel positions, i.e., it can accurately display the initial probability of fire points appearing at corresponding positions within the target location. Since the historical fire point multidimensional probability image is an image that displays the historical probability of fire points appearing at the corresponding positions of pixels in multiple dimensions, an accurate target fire point probability image can be obtained based on the target image, the historical fire point multidimensional probability image, and the initial fire point probability image. This achieves the goal of accurately determining the target probability of fire points appearing at the corresponding positions of pixels, thereby improving the technical effect of determining the accuracy of the determined fire point probability. This solves the technical problem of inaccurate fire point probability determination in related technologies. Attached Figure Description

[0016] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0017] Figure 1 This is a flowchart of a fire point determination method according to an embodiment of the present invention;

[0018] Figure 2 This is a structural block diagram of a fire point determination device according to an embodiment of the present invention. Detailed Implementation

[0019] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0020] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0021] Example 1

[0022] According to an embodiment of the present invention, an embodiment of a fire point determination method is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0023] First, it should be noted that the target image in this application can be an image taken by a satellite, and the target image can include visible light images and infrared images.

[0024] Figure 1 This is a flowchart of a fire point determination method according to an embodiment of the present invention, such as... Figure 1 As shown, the method includes the following steps:

[0025] Step S102: Obtain the target image of the target location and the historical fire point multi-dimensional probability image corresponding to the target location. The historical fire point multi-dimensional probability image is an image that displays the historical probability of fire points appearing in multiple dimensions at the corresponding positions of pixels in the image.

[0026] In the technical solution provided in step S102 of the present invention, satellite imagery of the target location can be acquired. The target imagery may include visible light images and infrared images. Based on historical data such as the location, frequency, and season of historical fires at the target location, a multi-dimensional probability image P of historical fire points is formed. i,j,t , where (i,j) represents the pixel position of the registered multi-source image, and t represents the current time. In this way, the target image of the target location, as well as the multi-dimensional probability image of the historical fire points corresponding to the target location, can be accurately obtained.

[0027] Step S104: Based on the target image, obtain the initial fire point probability image, wherein the initial fire point probability image is the image showing the initial probability of fire points appearing at the corresponding positions of pixels in the target image;

[0028] In the technical solution provided by step S104 of the present invention, since the target image can accurately reflect the current situation of the target location, the initial fire point probability image obtained based on the target image can more accurately display the initial probability of a fire point appearing at the corresponding position of the pixel, that is, it can more accurately display the initial probability of a fire point appearing at the corresponding position of the target location.

[0029] Step S106: Based on the target image, the historical fire point multi-dimensional probability image, and the initial fire point probability image, obtain the target fire point probability image, wherein the target fire point probability image is an image that displays the target probability of fire points appearing at the corresponding positions of pixels in the target image.

[0030] In the technical solution provided by step S106 of the present invention, since the historical fire point multi-dimensional probability image is an image that displays the historical probability of fire points appearing at the corresponding positions of pixels in multiple dimensions, and the initial fire point probability image can accurately display the initial probability of fire points appearing at the corresponding positions of pixels, an accurate target fire point probability image can be obtained based on the target image, the historical fire point multi-dimensional probability image and the initial fire point probability image, thereby achieving the purpose of accurately determining the target probability of fire points appearing at the corresponding positions of pixels.

[0031] Through steps S102 to S106, the process involves first acquiring a target image of the target location and a historical multi-dimensional probability image of fire points corresponding to the target location. The historical multi-dimensional probability image displays the historical probability of fire points appearing at corresponding pixel positions in multiple dimensions. Then, based on the target image, an initial fire point probability image is obtained, displaying the initial probability of fire points appearing at corresponding pixel positions in the target image. Finally, based on the target image, the historical multi-dimensional probability image, and the initial fire point probability image, a target fire point probability image is obtained, displaying the target probability of fire points appearing at corresponding pixel positions in the target image. Since the acquired target image is an image of the target location, it accurately reflects the current situation at the target location. Therefore, the initial fire point probability image obtained based on the target image can accurately display the initial probability of fire points appearing at corresponding pixel positions, i.e., it can accurately display the initial probability of fire points appearing at corresponding positions of the target location. Since the historical fire point multidimensional probability image is an image that displays the historical probability of fire points appearing at the corresponding positions of pixels in multiple dimensions, an accurate target fire point probability image can be obtained based on the target image, the historical fire point multidimensional probability image, and the initial fire point probability image. This achieves the goal of accurately determining the target probability of fire points appearing at the corresponding positions of pixels, thereby improving the technical effect of determining the accuracy of the determined fire point probability. This solves the technical problem of inaccurate fire point probability determination in related technologies.

[0032] As an optional embodiment, a target fire point probability image is obtained based on the target image, historical fire point multi-dimensional probability images, and an initial fire point probability image. This includes: inputting the target image, historical fire point multi-dimensional probability images, and the initial fire point probability image into a target fire point detection model to obtain the target fire point probability image. The target fire point detection model is obtained by training an initial fire point detection model based on sample pair data. The sample pair data includes positive sample images, historical positive sample fire point multi-dimensional probability images, initial positive sample fire point probability images, target positive sample fire point probability images, and negative sample images, historical negative sample fire point multi-dimensional probability images, initial negative sample fire point probability images, and target negative sample fire point probability images. The target positive sample fire point probability image is an image in which the probability of a fire point appearing at the corresponding position of a pixel in the image is greater than a first threshold. The target negative sample fire point probability image is an image in which the probability of a fire point appearing at the corresponding position of a pixel in the image is less than a second threshold. The first threshold is greater than the second threshold.

[0033] In this embodiment, the target fire detection model can be obtained by training an initial fire detection model with labeled data before use. For N << K, i.e., N is much smaller than K, where N is the number of labeled data and K is the number of target images. A self-supervised task based on contrastive learning can be constructed based on the initial fire probability image to train the initial fire detection model. Specifically, based on the initial fire probability image, the top 5% of pixels with a relatively high initial probability of fire at the corresponding position (equivalent to the initial positive sample fire probability image mentioned above) and the bottom 5% of pixels with a relatively low initial probability of fire at the corresponding position (equivalent to the initial negative sample fire probability image mentioned above) can be selected to form a sample set. In this way, the target fire detection model can be obtained by training the initial fire detection model with a small amount of annotation, and a target fire detection model with good performance and high accuracy can be obtained. Therefore, by inputting the target image, the historical fire multi-dimensional probability image, and the initial fire probability image into the target fire detection model, an accurate target fire probability image can be obtained.

[0034] It should be noted that when training the initial fire detection model to obtain the target fire detection model based on sample pairs of data, the following method can be used: Concatenated convolution, nonlinear transformation, and downsampling operations are performed on the sample pairs of data to form an encoding structure in the initial fire detection model, resulting in the first fire detection model. Then, convolution, nonlinear transformation, and upsampling operations are performed, and simultaneously concatenated with the corresponding downsampled data to form a decoding structure in the first fire detection model, resulting in the second fire detection model. It is ensured that the encoding layer (equivalent to the encoding structure above) and decoding layer (equivalent to the decoding structure above) in the second fire detection model are consistent. Simultaneously, by setting a tape measure parameter, the output data space dimension of the encoding layer and the corresponding decoding layer are made consistent, and the feature dimension of the last layer of the decoding layer is one, thus obtaining the target fire detection model.

[0035] It should also be noted that when inputting the target image, historical fire point multi-dimensional probability image, and initial fire point probability image into the target fire point detection model to obtain the target fire point probability image, the historical fire point multi-dimensional probability image and the initial fire point probability image can be superimposed on the fully connected layer of the target fire point detection model. Inputting the target image into the target fire point detection model after superimposing the fully connected layer can obtain the target fire point probability image.

[0036] As an optional embodiment, before inputting the target image, historical fire point multi-dimensional probability image, and initial fire point probability image into the target fire point detection model to obtain the target fire point probability image, the method further includes: obtaining the initial fire point detection model and constructing a loss function for model training, wherein the loss function includes a cross-entropy loss function; based on the loss function, the initial fire point detection model is trained using sample pair data to obtain the target fire point detection model.

[0037] In this embodiment, based on the loss function, the initial fire detection model is trained using sample pairs of data to obtain an accurate target fire detection model.

[0038] As an optional embodiment, the target fire point probability image is obtained based on the target image, the historical fire point multi-dimensional probability image, and the initial fire point probability image, including: determining the probability clustering result of fire points appearing at the corresponding positions of pixels in the target image based on the historical fire point multi-dimensional probability image and the initial fire point probability image; and obtaining the target fire point probability image based on the probability clustering result of fire points appearing at the corresponding positions of pixels in the target image.

[0039] In this embodiment, a clustering method, such as k-means clustering, can be used to divide the probability of a fire point appearing at the corresponding position of a pixel in the target image into T classes, where T can be 10. The cluster center of each class is the new sample primitive. Normalized linear weight parameters {w} are constructed by random generation. t |t∈(1,T)}, a new sample is obtained by weighted summation of the sample primitives. Corresponding linear weight w t This serves as the soft label for the new sample classification. These labels form new self-supervised sample pairs, and based on the clustering results of the probability of fire points appearing at corresponding positions of pixels in the target image, an accurate target fire point probability image is obtained.

[0040] It should be noted that the dimensions of the fully connected layers in the initial fire detection model can be modified to match the clustering category T. Based on the sample pairs of data, the parameters of the initial fire detection model are iteratively learned using gradient calculation methods until the parameters of the convolutional network converge, at which point the iteration stops. Through comparative learning, the parameters of the convolutional network are obtained, thus yielding the target fire detection model.

[0041] As an optional embodiment, before acquiring the target image of the target location, the method further includes: acquiring an initial image of the target location; registering the initial image according to predetermined parameter information to obtain the target image, wherein the predetermined parameter information includes a predetermined resolution and spatial corresponding position information of each vertex of the image.

[0042] In this embodiment, the initial image can be spatially calibrated based on predetermined parameter information to ensure that each image in the resulting target image has the same observation position, viewing angle, and resolution. Based on this, a global normalization operation is performed on images from different sources within the target image to improve image contrast and facilitate subsequent calculations. The processed image can be represented as... L={L l|l∈(1,N)}, where I represents the dataset corresponding to the target image, C represents the number of channels in the target image (e.g., if the target image includes visible light and infrared images, there are 3 channels for visible light and 1 channel for infrared, so C is 4), K represents the number of target images, L represents the positions where the marked pixels have a higher probability of containing fire points, and N is the number of marked data points, where N << K, meaning N is much smaller than K. In this way, an accurate registered target image can be quickly obtained based on a small number of annotations.

[0043] As an optional embodiment, obtaining a multi-dimensional probability image of historical fire points corresponding to a target location includes: obtaining historical fire point data of the target location; and determining a multi-dimensional probability image of historical fire points corresponding to the target location based on the historical fire point data, wherein the number of dimensions is determined according to the number of types of historical fire point data.

[0044] In this embodiment, the type and quantity of historical fire point data can be determined according to actual needs, and the required historical fire point data for the target location can be obtained. Based on the historical fire point data, an accurate multi-dimensional probability image of historical fire points corresponding to the target location can be quickly determined.

[0045] As an optional embodiment, obtaining an initial fire point probability image based on the target image includes: determining the initial fire point probability image based on the target image and the target algorithm, wherein the target algorithm includes the Constrained Energy Minimization (CEM) algorithm.

[0046] In this embodiment, the data corresponding to the target image can be two-dimensionalized and normalized. Then, the autocorrelation matrix R is calculated based on the autocorrelation function. Next, the pixels with the highest brightness values ​​in the target image are selected as initial fire points. Simultaneously, the initial fire points are labeled in the target image to obtain an initial fire point dataset. Then, clustering, such as k-means clustering, is used to determine the cluster center positions. The value d corresponding to the cluster center position is used as the baseline fire point. Based on this, the linear filter parameters are calculated according to the following formula:

[0047]

[0048] Finally, based on the linear filter parameters w mentioned above... CEM The probability of a fire point appearing at each pixel location in the target image is obtained by linearly weighting and summing the values. Then, an initial fire point probability image is obtained by applying an automatic thresholding method or by using a 1% threshold. In this way, an accurate initial fire probability image can be obtained quickly.

[0049] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the present invention is not limited to the described order of actions, because according to the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0050] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0051] Example 2

[0052] According to an embodiment of the present invention, an apparatus for implementing the above-described fire point determination method is also provided. Figure 2 This is a structural block diagram of a fire point determining device according to an embodiment of the present invention, such as... Figure 2 As shown, the device includes a first acquisition module 202, a second acquisition module 204, and a third acquisition module 206. The device will be described in detail below.

[0053] The first acquisition module 202 is used to acquire a target image of the target location and a historical fire point multi-dimensional probability image corresponding to the target location, wherein the historical fire point multi-dimensional probability image is an image that displays the historical probability of fire points appearing at the corresponding positions of pixels in the image in multiple dimensions; the second acquisition module 204 is connected to the first acquisition module 202 and is used to obtain an initial fire point probability image based on the target image, wherein the initial fire point probability image is an image that displays the initial probability of fire points appearing at the corresponding positions of pixels in the target image; the third acquisition module 206 is connected to the second acquisition module 204 and is used to obtain a target fire point probability image based on the target image, the historical fire point multi-dimensional probability image, and the initial fire point probability image, wherein the target fire point probability image is an image that displays the target probability of fire points appearing at the corresponding positions of pixels in the target image.

[0054] It should be noted that the first acquisition module 202, the second acquisition module 204 and the third acquisition module 206 mentioned above correspond to steps S102 to S106 in the implementation of the fire point determination method. The multiple modules and the corresponding steps implement the same instances and application scenarios, but are not limited to the content disclosed in the above embodiment 1.

[0055] Example 3

[0056] According to another aspect of the present invention, an electronic device is also provided, comprising: a processor; and a memory for storing processor-executable instructions, wherein the processor is configured to execute instructions to implement the fire point determination method of any of the above embodiments.

[0057] Example 4

[0058] According to another aspect of the present invention, a computer-readable storage medium is also provided, which, when the instructions in the computer-readable storage medium are executed by a processor of an electronic device, enables the electronic device to perform the fire point determination method described above.

[0059] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0060] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0061] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0062] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0063] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0064] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0065] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for determining a fire point, characterized in that, include: Acquire a target image of the target location and a historical fire point multi-dimensional probability image corresponding to the target location. The historical fire point multi-dimensional probability image is an image that displays the historical probability of fire points occurring in multiple dimensions at the corresponding positions of pixels in the image. The historical fire point multi-dimensional probability image is obtained based on the historical data of the target location. The historical data includes the historical fire point location, the number of occurrences, and the season of occurrence. Based on the target image, an initial fire point probability image is obtained, wherein the initial fire point probability image is an image that displays the initial probability of a fire point appearing at the corresponding position of a pixel in the target image; Based on the target image, the historical fire point multi-dimensional probability image, and the initial fire point probability image, a target fire point probability image is obtained, wherein the target fire point probability image is an image that displays the target probability of fire points appearing at the corresponding positions of pixels in the target image; The step of obtaining an initial fire point probability image based on the target image includes: performing two-dimensionalization and normalization on the data corresponding to the target image, and then calculating the autocorrelation matrix using an autocorrelation function. The initial fire point data set is obtained by selecting the top 1% of pixels with the highest brightness values ​​in the target image; the cluster center positions of the initial fire point data set are determined by k-means clustering, and the values ​​corresponding to the cluster center positions are used as the initial fire points. As a reference firing point; based on the linear filter parameters The probability of a fire point appearing at the location corresponding to each pixel in the initial fire point probability image is obtained by performing a linear weighted summation on each pixel in the target image, wherein the linear filter parameters are calculated according to the following formula: 。 2. The method according to claim 1, characterized in that, The step of obtaining the target fire point probability image based on the target image, the historical fire point multi-dimensional probability image, and the initial fire point probability image includes: The target image, the historical fire point multi-dimensional probability image, and the initial fire point probability image are input into the target fire point detection model to obtain the target fire point probability image. The target fire point detection model is obtained by training the initial fire point detection model based on sample pair data. The sample pair data includes positive sample images, historical positive sample fire point multi-dimensional probability images, initial positive sample fire point probability images, target positive sample fire point probability images, and negative sample images, historical negative sample fire point multi-dimensional probability images, initial negative sample fire point probability images, and target negative sample fire point probability images. The target positive sample fire point probability image is an image in which the probability of a fire point appearing at the corresponding position of a pixel in the image is greater than a first threshold. The target negative sample fire point probability image is an image in which the probability of a fire point appearing at the corresponding position of a pixel in the image is less than a second threshold. The first threshold is greater than the second threshold.

3. The method according to claim 2, characterized in that, Before inputting the target image, the historical fire point multi-dimensional probability image, and the initial fire point probability image into the target fire point detection model to obtain the target fire point probability image, the method further includes: Obtain an initial fire detection model and construct a loss function for model training, wherein the loss function includes a cross-entropy loss function; Based on the loss function, the initial fire detection model is trained using the sample pair data to obtain the target fire detection model.

4. The method according to claim 1, characterized in that, The step of obtaining the target fire point probability image based on the target image, the historical fire point multi-dimensional probability image, and the initial fire point probability image includes: Based on the historical fire point multidimensional probability image and the initial fire point probability image, determine the probability clustering result of fire points appearing at the corresponding positions of pixels in the target image; Based on the clustering results of the probability of fire points appearing at the corresponding positions of pixels in the target image, the target fire point probability image is obtained.

5. The method according to claim 1, characterized in that, Before acquiring the target image of the target location, the process also includes: Obtain an initial image of the target location; The initial image is registered according to predetermined parameter information to obtain the target image, wherein the predetermined parameter information includes a predetermined resolution and spatial position information of each vertex of the image.

6. The method according to claim 1, characterized in that, Obtaining a multi-dimensional probability image of historical fire points corresponding to the target location, including: Obtain historical fire point data for the target location; Based on the historical fire point data, a multi-dimensional probability image of the historical fire point corresponding to the target location is determined, wherein the number of dimensions of the multi-dimensional image is determined according to the number of types of the historical fire point data.

7. A fire point determination device, characterized in that, include: The first acquisition module is used to acquire a target image of the target location and a historical fire point multi-dimensional probability image corresponding to the target location. The historical fire point multi-dimensional probability image is an image that displays the historical probability of fire points appearing in multiple dimensions at the corresponding positions of pixels in the image. The historical fire point multi-dimensional probability image is obtained based on the historical data of the target location. The historical data includes the historical fire point location, the number of occurrences, and the season of occurrence. The second acquisition module is used to obtain an initial fire point probability image based on the target image, wherein the initial fire point probability image is an image that displays the initial probability of a fire point appearing at the corresponding position of a pixel in the target image; The third acquisition module is used to obtain a target fire point probability image based on the target image, the historical fire point multi-dimensional probability image and the initial fire point probability image, wherein the target fire point probability image is an image that displays the target probability of fire points appearing at the corresponding positions of pixels in the target image; The step of obtaining an initial fire point probability image based on the target image includes: performing two-dimensionalization and normalization on the data corresponding to the target image, and then calculating the autocorrelation matrix using an autocorrelation function. The initial fire point data set is obtained by selecting the top 1% of pixels with the highest brightness values ​​in the target image; the cluster center positions of the initial fire point data set are determined by k-means clustering, and the values ​​corresponding to the cluster center positions are used as the initial fire points. As a reference firing point; based on the linear filter parameters The probability of a fire point appearing at the location corresponding to each pixel in the initial fire point probability image is obtained by performing a linear weighted summation on each pixel in the target image, wherein the linear filter parameters are calculated according to the following formula: 。 8. An electronic device, characterized in that, include: processor; Memory used to store the processor's executable instructions; The processor is configured to execute the instructions to implement the fire point determination method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, When the instructions in the computer-readable storage medium are executed by the processor of the electronic device, the electronic device is able to perform the fire point determination method as described in any one of claims 1 to 6.

Images