Disclosure of Invention
The invention provides a method and a system for processing data of remote sensing monitoring natural disasters.
In a first aspect, an embodiment of the present invention provides a data processing method for remote sensing monitoring of natural disasters, which is applied to a computer device, where the computer device is in communication connection with a remote sensing device, and the method includes:
acquiring known detection points acquired by remote sensing equipment in a current acquisition stage;
determining an original remote sensing image based on known detection points in the current acquisition stage;
carrying out remote sensing information processing on the original remote sensing image through a preset neural network algorithm to obtain a remote sensing image to be analyzed;
determining a first mapping matrix between detection indexes of a remote sensing image to be analyzed and detection indexes of a historical disaster remote sensing image, wherein the first mapping matrix comprises the following steps: respectively constructing a logical relationship tree of each detection point in a remote sensing image to be analyzed and a historical disaster remote sensing image to obtain a plurality of first remote sensing characteristic locus diagrams and a plurality of second remote sensing characteristic locus diagrams, determining a first matching confidence coefficient between each first remote sensing characteristic locus diagram and each second remote sensing characteristic locus diagram to obtain a first mapping matrix, wherein the historical disaster remote sensing image is determined based on disaster data at a sample stage;
extracting candidate detection points meeting preset association rules of the remote sensing image to be analyzed from the historical disaster remote sensing image to obtain a first candidate detection point set;
determining a first target detection point from the first candidate detection point set based on the first mapping matrix, and adjusting detection point information of the remote sensing image to be analyzed according to the first target detection point to obtain a target disaster remote sensing image;
determining a second mapping matrix between the detection index of the target disaster remote sensing image and the detection index of the remote sensing image to be analyzed, wherein the second mapping matrix comprises the following steps: respectively constructing a logical relationship tree of each detection point in the remote sensing image to be analyzed and the target disaster remote sensing image to obtain a plurality of third remote sensing characteristic locus diagrams and a plurality of fourth remote sensing characteristic locus diagrams, determining a second matching confidence coefficient between each third remote sensing characteristic locus diagram and each fourth remote sensing characteristic locus diagram to obtain a second mapping matrix;
and carrying out detection point prediction on the remote sensing image to be analyzed according to the second mapping matrix and the detection points in the remote sensing image to be analyzed to obtain a disaster prediction result.
Optionally, the detection index of the remote sensing image to be analyzed includes: the disaster trend index of each known detection point in the remote sensing image to be analyzed and the disaster weight index in the remote sensing image to be analyzed are obtained; the detection indexes of the remote sensing image of the historical disaster comprise: the disaster trend index of each detection point in the historical disaster remote sensing image and the disaster weight index in the historical disaster remote sensing image are obtained;
determining a first target detection point from the first candidate detection point set based on the first mapping matrix, and adjusting detection point information of the remote sensing image to be analyzed according to the first target detection point, wherein the step comprises the following steps:
determining a first mapping function of each detection point in the remote sensing image to be analyzed relative to each detection point in the historical disaster remote sensing image based on the first matching confidence, wherein the first mapping function is used for reflecting the attention degree of each detection point in the remote sensing image to be analyzed to each detection point in the historical disaster remote sensing image;
determining corresponding candidate detection points from the first candidate detection point set according to the sequence of the first mapping function from high to low, wherein the corresponding candidate detection points serve as first target detection points;
and generating corresponding detection points at corresponding positions in the remote sensing image to be analyzed according to the disaster weight index of the first target detection points in the historical disaster remote sensing image and the disaster trend index corresponding to the first target points.
Optionally, the step of predicting the detection points of the remote sensing image to be analyzed according to the second mapping matrix and the detection points in the remote sensing image to be analyzed to obtain a disaster prediction result includes:
acquiring detection points, and acquiring corresponding second mapping matrixes according to the detection points;
based on detection points, obtaining a detection feature set through a first feature extraction network included in a disaster prediction model, wherein the detection feature set comprises a preset number of detection features;
acquiring a reference feature set through a second feature extraction network included in the disaster prediction model based on a second mapping matrix, wherein the reference feature set includes a preset number of reference features;
based on the detection feature set, according to each frame of detection features in the detection feature set, obtaining a first maximum pooling feature value through a maximum pooling layer included in a first attention mechanism, wherein the first attention mechanism belongs to a disaster prediction model;
according to each frame of detection features in the detection feature set, a first average pooling feature value is obtained through a global average pooling layer included in a first attention mechanism;
acquiring a first fusion feature through a convolutional layer included in a first attention mechanism based on a first maximum pooling feature value and a first average pooling feature value according to each frame detection feature in a detection feature set;
according to each frame of detection features in the detection feature set, based on the first fusion features and the detection features, obtaining first feature vectors through a first global average pooling layer included in a first attention mechanism, wherein each first feature vector corresponds to one detection feature;
acquiring a second maximum pooling characteristic value through a maximum pooling layer included in a second attention mechanism according to each frame of reference feature in the reference feature set, wherein the second attention mechanism belongs to a disaster prediction model;
acquiring a second average pooling characteristic value through a global average pooling layer included in a second attention mechanism according to each frame of reference characteristic in the reference characteristic set;
acquiring a second fusion feature through a convolutional layer included in a second attention mechanism based on a second maximum pooling feature value and a second average pooling feature value according to each frame of reference feature in the reference feature set;
acquiring second feature vectors through a second global average pooling layer included in a second attention mechanism according to each frame of reference feature in the reference feature set based on the second fusion feature and the reference feature, wherein each second feature vector corresponds to one reference feature;
fusing a preset number of first eigenvectors and a preset number of second eigenvectors to obtain a preset number of target eigenvectors, wherein each target eigenvector comprises a first eigenvector and a second eigenvector;
acquiring disaster matching scores corresponding to detection points through a full-connection layer included in a disaster prediction model based on a preset number of target feature vectors;
and determining a disaster prediction result of the remote sensing image to be analyzed according to the disaster matching score.
Optionally, the step of obtaining a disaster matching score corresponding to the detection point through a full-link layer included in the disaster prediction model based on a preset number of target feature vectors includes:
acquiring target prediction vectors through an attention mechanism included in a disaster prediction model based on a preset number of target feature vectors, wherein the target prediction vectors are determined according to the preset number of target feature vectors and a preset number of reference coefficients, and each target feature vector corresponds to one preset reference coefficient;
and acquiring disaster matching scores corresponding to the detection points through a full-connection layer included in the disaster prediction model based on the target prediction vector.
Optionally, the step of performing remote sensing information processing on the original remote sensing image through a preset neural network algorithm to obtain a remote sensing image to be analyzed includes:
constructing a logic relation tree of known detection points in an original remote sensing image;
determining matching confidence between every two known detection points in the original remote sensing image according to the logical relationship tree of the known detection points;
determining a third mapping function of each known detection point in the original remote sensing image relative to other known detection points based on the matching confidence degree between every two known detection points, wherein the third mapping function is used for reflecting the attention degree of each detection point in the original remote sensing image to other known detection points in the track of the detection point;
and adjusting the logical relation tree of the known detection points in the original remote sensing image according to the third mapping function to obtain the remote sensing image to be analyzed.
Optionally, the step of determining the historical disaster remote sensing image based on the disaster data at the sample stage includes:
acquiring disaster data of a sample stage;
constructing a plurality of sample disaster remote sensing images according to the appointed acquisition stage and disaster data;
integrating the plurality of sample disaster remote sensing images according to time, determining detection points with highest occurrence frequency at the same moment from the plurality of integrated sample disaster remote sensing images, and constructing and obtaining a target sample disaster remote sensing image according to the detection points with highest occurrence frequency at the same moment;
constructing a logical relation tree of each detection point in the target sample disaster remote sensing image;
determining a matching confidence coefficient between every two detection points in the target sample disaster remote sensing image according to the logic relation tree of each detection point;
determining a fourth mapping function of each detection point in the target sample disaster remote sensing image relative to other detection points based on the matching confidence coefficient between every two detection points, wherein the fourth mapping function is used for reflecting the attention degree of each detection point in the target sample disaster remote sensing image to other detection points in the track of the detection point;
and adjusting the logical relation tree of the detection points in the target sample disaster remote sensing image according to the fourth mapping function to obtain the historical disaster remote sensing image.
Optionally, the computer device is further in communication connection with a disaster processing server, and adjusts detection point information of the remote sensing image to be analyzed according to the first target detection point to obtain a target disaster remote sensing image, including:
acquiring a first target detection point, detection point information corresponding to the first target detection point and a detection feature vector of the detection point information, wherein the detection feature vector of the detection point information comprises first disaster data and at least one second disaster data of the first disaster data;
determining a data segment field configured in the detection point information according to the first preset weight and the first preset disaster type as first disaster data;
determining at least one data segment field configured along the edge of the first disaster data in the detection point information according to a second preset weight and a second preset disaster type as second disaster data of the first disaster data;
searching a target disaster type matched with the first target detection point;
creating a marking task for the first target detection point, wherein the marking task comprises an identifier of the first target detection point;
pushing the marking task to a target disaster processing server;
receiving a marking request sent by a target disaster processing server for executing a marking task, wherein the marking request carries an identifier of a first target detection point;
returning the detection point information corresponding to the first target detection point and the detection feature vector of the detection point information to the target disaster processing server according to the identifier of the first target detection point, so that the target disaster processing server marks the first disaster data or the second disaster data of each detection feature vector of the detection point information to obtain a marked feature vector of the detection point information, wherein the marked feature vector of the detection point information comprises a marked object in the detection feature vector and a first sequence of the marked object in the detection feature vector;
receiving a marked feature vector of detection point information submitted by a target disaster processing server;
converting a first sequence of the marked object in the detection feature vector to obtain a second sequence of the marked object in the first target detection point;
determining a marking field in the first target detection point according to the second sequence of the marking object in the first target detection point;
determining a second sequence of the mark object in the first target detection point and a mark field in the first target detection point as a mark feature vector of the first target detection point;
and determining a target disaster remote sensing image from the remote sensing image to be analyzed according to the marking characteristic vector of the detection point information and the marking characteristic vector of the first target detection point.
In a second aspect, an embodiment of the present invention provides a data processing system for remote sensing monitoring of natural disasters, which is applied to a computer device, where the computer device is in communication connection with a remote sensing device, and the system includes:
the acquisition module is used for acquiring known detection points acquired by the remote sensing equipment in the current acquisition stage; determining an original remote sensing image based on known detection points in the current acquisition stage; processing remote sensing information of the original remote sensing image through a preset neural network algorithm to obtain a remote sensing image to be analyzed;
the determining module is used for determining a first mapping matrix between the detection index of the remote sensing image to be analyzed and the detection index of the historical disaster remote sensing image, and comprises the following steps: respectively constructing a logical relationship tree of each detection point in a remote sensing image to be analyzed and a historical disaster remote sensing image to obtain a plurality of first remote sensing characteristic locus diagrams and a plurality of second remote sensing characteristic locus diagrams, determining a first matching confidence coefficient between each first remote sensing characteristic locus diagram and each second remote sensing characteristic locus diagram to obtain a first mapping matrix, wherein the historical disaster remote sensing image is determined based on disaster data at a sample stage; extracting candidate detection points meeting preset association rules of the remote sensing image to be analyzed from the historical disaster remote sensing image to obtain a first candidate detection point set; determining a first target detection point from the first candidate detection point set based on the first mapping matrix, and adjusting detection point information of the remote sensing image to be analyzed according to the first target detection point to obtain a target disaster remote sensing image; determining a second mapping matrix between the detection index of the target disaster remote sensing image and the detection index of the remote sensing image to be analyzed, wherein the second mapping matrix comprises the following steps: respectively constructing a logical relationship tree of each detection point in the remote sensing image to be analyzed and the target disaster remote sensing image to obtain a plurality of third remote sensing characteristic locus diagrams and a plurality of fourth remote sensing characteristic locus diagrams, determining a second matching confidence coefficient between each third remote sensing characteristic locus diagram and each fourth remote sensing characteristic locus diagram to obtain a second mapping matrix;
and the prediction module is used for predicting the detection points of the remote sensing image to be analyzed according to the second mapping matrix and the detection points in the remote sensing image to be analyzed to obtain a disaster prediction result.
Optionally, the detection index of the remote sensing image to be analyzed includes: the disaster trend index of each known detection point in the remote sensing image to be analyzed and the disaster weight index in the remote sensing image to be analyzed are obtained; the detection indexes of the remote sensing image of the historical disaster comprise: the disaster trend index of each detection point in the historical disaster remote sensing image and the disaster weight index in the historical disaster remote sensing image are obtained;
the determination module is specifically configured to:
determining a first mapping function of each detection point in the remote sensing image to be analyzed relative to each detection point in the historical disaster remote sensing image based on the first matching confidence, wherein the first mapping function is used for reflecting the attention degree of each detection point in the remote sensing image to be analyzed to each detection point in the historical disaster remote sensing image; determining corresponding candidate detection points from the first candidate detection point set according to the sequence of the first mapping function from high to low, wherein the corresponding candidate detection points serve as first target detection points; and generating corresponding detection points at corresponding positions in the remote sensing image to be analyzed according to the disaster weight index of the first target detection points in the historical disaster remote sensing image and the disaster trend index corresponding to the first target points.
Optionally, the prediction module is specifically configured to:
acquiring detection points, and acquiring corresponding second mapping matrixes according to the detection points; based on detection points, obtaining a detection feature set through a first feature extraction network included in a disaster prediction model, wherein the detection feature set comprises a preset number of detection features; acquiring a reference feature set through a second feature extraction network included in the disaster prediction model based on a second mapping matrix, wherein the reference feature set includes a preset number of reference features; based on the detection feature set, according to each frame of detection features in the detection feature set, obtaining a first maximum pooling feature value through a maximum pooling layer included in a first attention mechanism, wherein the first attention mechanism belongs to a disaster prediction model; according to each frame of detection features in the detection feature set, a first average pooling feature value is obtained through a global average pooling layer included in a first attention mechanism; acquiring a first fusion feature through a convolutional layer included in a first attention mechanism based on a first maximum pooling feature value and a first average pooling feature value according to each frame detection feature in a detection feature set; according to each frame of detection features in the detection feature set, based on the first fusion features and the detection features, obtaining first feature vectors through a first global average pooling layer included in a first attention mechanism, wherein each first feature vector corresponds to one detection feature; acquiring a second maximum pooling characteristic value through a maximum pooling layer included in a second attention mechanism according to each frame of reference feature in the reference feature set, wherein the second attention mechanism belongs to a disaster prediction model; acquiring a second average pooling characteristic value through a global average pooling layer included in a second attention mechanism according to each frame of reference characteristic in the reference characteristic set; acquiring a second fusion feature through a convolutional layer included in a second attention mechanism based on a second maximum pooling feature value and a second average pooling feature value according to each frame of reference feature in the reference feature set; acquiring second feature vectors through a second global average pooling layer included in a second attention mechanism according to each frame of reference feature in the reference feature set based on the second fusion feature and the reference feature, wherein each second feature vector corresponds to one reference feature; fusing a preset number of first eigenvectors and a preset number of second eigenvectors to obtain a preset number of target eigenvectors, wherein each target eigenvector comprises a first eigenvector and a second eigenvector; acquiring disaster matching scores corresponding to detection points through a full-connection layer included in a disaster prediction model based on a preset number of target feature vectors; and determining a disaster prediction result of the remote sensing image to be analyzed according to the disaster matching score.
Compared with the prior art, the beneficial effects provided by the invention comprise: the invention discloses a method and a system for processing data of remote sensing monitoring natural disasters, which are characterized in that known detection points acquired by remote sensing equipment in a current acquisition stage are obtained; further determining an original remote sensing image and a remote sensing image to be analyzed; by determining the remote sensing image to be analyzed and the historical disaster remote sensing image, obtaining the target disaster remote sensing image, and determining the second mapping matrix between the detection index of the target disaster remote sensing image and the detection index of the remote sensing image to be analyzed, the detection point prediction can be carried out on the remote sensing image to be analyzed according to the second mapping matrix and the detection point in the remote sensing image to be analyzed, so that the disaster prediction result obtained through prediction can be determined based on the remote sensing image to be analyzed and the historical disaster remote sensing image, and the disaster prediction result with more credible reference degree can be obtained.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention. It is to be understood that the embodiments described are only a few embodiments of the present invention, and not all embodiments. The components of embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.
Thus, the following detailed description of the embodiments of the present invention, presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that: like reference numbers and letters refer to like items in the following figures, and thus, once an item is defined in one figure, it need not be further defined or explained in subsequent figures.
In the description of the present invention, it is to be understood that the terms "upper", "lower", "inner", "outer", "left", "right", and the like indicate orientations or positional relationships based on orientations or positional relationships shown in the drawings, or orientations or positional relationships customarily placed in use of products of this application, or orientations or positional relationships customarily understood by those skilled in the art, merely for the convenience of description and simplification of the description, and are not intended to indicate or imply that the referenced devices or elements must have a particular orientation, be constructed in a particular orientation, and be operated, and therefore, are not to be construed as limiting the present invention.
Furthermore, the terms "first," "second," and the like are used merely to distinguish one description from another, and are not to be construed as indicating or implying relative importance.
In the description of the present invention, it is also to be noted that, unless otherwise explicitly stated or limited, the terms "disposed" and "connected" are to be interpreted broadly, and for example, "connected" may be a fixed connection, a detachable connection, or an integral connection; can be mechanically or electrically connected; the connection may be direct or indirect via an intermediate medium, and may be a communication between the two elements. The specific meanings of the above terms in the present invention can be understood by those skilled in the art according to specific situations.
The following detailed description of embodiments of the invention refers to the accompanying drawings.
Fig. 1 is an interaction diagram of a data processing system for remotely monitoring natural disasters according to an embodiment of the present disclosure. A data processing system for remotely monitoring natural disasters may include a computer device 100 and a remote sensing device 200 communicatively coupled to the computer device 100. The remote monitoring natural disaster data processing system shown in fig. 1 is only one possible example, and in other possible embodiments, the remote monitoring natural disaster data processing system may include only a portion of the components shown in fig. 1 or may also include other components.
In this embodiment, the computer device 100 and the remote sensing device 200 in the data processing of remote sensing monitoring natural disasters may cooperatively execute the data processing method of remote sensing monitoring natural disasters described in the following method embodiments, and specific steps executed by the computer device 100 and the remote sensing device 200 may refer to the detailed description of the following method embodiments.
To solve the technical problem in the foregoing background, fig. 2 is a schematic flow chart of a method for processing data of remote sensing monitoring natural disasters according to an embodiment of the present disclosure, where the method for processing data of remote sensing monitoring natural disasters according to the present embodiment may be executed by the computer device 100 shown in fig. 1, and the method for processing data of remote sensing monitoring natural disasters is described in detail below.
Step 201, obtaining known detection points acquired by the remote sensing equipment 200 in the current acquisition stage.
Step 202, determining an original remote sensing image based on known detection points in the current acquisition phase.
And 203, carrying out remote sensing information processing on the original remote sensing image through a preset neural network algorithm to obtain a remote sensing image to be analyzed.
And 204, determining a first mapping matrix between the detection indexes of the remote sensing image to be analyzed and the detection indexes of the remote sensing image of the historical disaster.
The method comprises the following steps: respectively constructing a logical relationship tree of each detection point in a remote sensing image to be analyzed and a historical disaster remote sensing image to obtain a plurality of first remote sensing characteristic locus diagrams and a plurality of second remote sensing characteristic locus diagrams, determining a first matching confidence coefficient between each first remote sensing characteristic locus diagram and each second remote sensing characteristic locus diagram to obtain a first mapping matrix, wherein the historical disaster remote sensing image is determined based on disaster data at a sample stage.
Step 205, extracting candidate detection points meeting preset association rules of the remote sensing images to be analyzed from the historical disaster remote sensing images to obtain a first candidate detection point set.
And step 206, determining a first target detection point from the first candidate detection point set based on the first mapping matrix, and adjusting detection point information of the remote sensing image to be analyzed according to the first target detection point to obtain a target disaster remote sensing image.
And step 207, determining a second mapping matrix between the detection index of the target disaster remote sensing image and the detection index of the remote sensing image to be analyzed.
The method comprises the following steps: and respectively constructing a logical relationship tree of each detection point in the remote sensing image to be analyzed and the target disaster remote sensing image to obtain a plurality of third remote sensing characteristic locus diagrams and a plurality of fourth remote sensing characteristic locus diagrams, determining a second matching confidence coefficient between each third remote sensing characteristic locus diagram and each fourth remote sensing characteristic locus diagram, and obtaining a second mapping matrix.
And 208, carrying out detection point prediction on the remote sensing image to be analyzed according to the second mapping matrix and the detection points in the remote sensing image to be analyzed to obtain a disaster prediction result.
In the embodiment of the invention, the preset neural network algorithm can be a self-attention mechanism, and through the steps, a disaster prediction result with a large reference value can be obtained.
On the basis, the detection indexes of the remote sensing image to be analyzed comprise: the disaster trend index of each known detection point in the remote sensing image to be analyzed and the disaster weight index in the remote sensing image to be analyzed are obtained; the detection indexes of the remote sensing image of the historical disaster comprise: the disaster trend index of each detection point in the historical disaster remote sensing image and the disaster weight index in the historical disaster remote sensing image are obtained;
as an alternative embodiment, the foregoing step 206 may be implemented by the following detailed description.
And a substep 206-1 of determining a first mapping function of each detection point in the remote sensing image to be analyzed with respect to each detection point in the remote sensing image of the historical disaster based on the first matching confidence.
The first mapping function is used for reflecting the attention degree of each detection point in the remote sensing image to be analyzed to each detection point in the historical disaster remote sensing image.
Sub-step 206-2, determining corresponding candidate detection points from the first set of candidate detection points as first target detection points in descending order according to the first mapping function.
And a substep 206-3 of generating corresponding detection points at corresponding positions in the remote sensing image to be analyzed according to the disaster weight index of the first target detection points in the historical disaster remote sensing image and the disaster trend index corresponding to the first target points.
On this basis, in order to more clearly describe the scheme of the present invention, a detailed implementation step of the foregoing step 208 is provided below.
And substep 208-1, obtaining the detection points and obtaining corresponding second mapping matrices according to the detection points.
And a substep 208-2 of acquiring a set of detection features through a first feature extraction network included in the disaster prediction model based on the detection points.
Wherein the detection feature set comprises a preset number of detection features.
And a substep 208-3 of obtaining a reference feature set through a second feature extraction network included in the disaster prediction model based on the second mapping matrix.
Wherein the reference feature set comprises a preset number of reference features.
Sub-step 208-4, based on the detected feature set, obtaining a first maximum pooled feature value from the maximum pooled layer included in the first attention mechanism according to each frame of detected features in the detected feature set.
Wherein, the first attention mechanism belongs to a disaster prediction model.
Sub-step 208-5, obtaining a first average pooled feature value by a global average pooling layer comprised by the first attention mechanism, based on each frame of detected features in the set of detected features.
Sub-step 208-6, obtaining a first fused feature through the convolutional layer included in the first attention mechanism based on the first maximum pooled feature value and the first average pooled feature value according to each frame of the detected features in the detected feature set.
Sub-step 208-7, obtaining first feature vectors through a first global average pooling layer comprised by the first attention mechanism based on the first fused features and the detected features according to each frame of detected features in the detected feature set, wherein each first feature vector corresponds to one detected feature.
Sub-step 208-8, obtaining a second maximum pooled feature value by the maximum pooling layer comprised by the second attention mechanism, from each frame of reference feature in the set of reference features.
Wherein the second attention mechanism belongs to a disaster prediction model.
Sub-step 208-9, obtaining a second average pooled feature value by a global average pooling layer comprised by the second attention mechanism, based on each frame of reference feature in the set of reference features.
Sub-step 208-10, obtaining a second fusion feature through the convolutional layer comprised by the second attention mechanism based on the second maximum pooling feature value and the second average pooling feature value according to each frame of reference feature in the reference feature set.
Sub-step 208-11, obtaining a second feature vector through a second global average pooling layer included in the second attention mechanism based on the second fused feature and the reference feature according to each frame of reference feature in the reference feature set.
Wherein each second feature vector corresponds to a reference feature.
And a substep 208-12 of fusing the preset number of first eigenvectors and the preset number of second eigenvectors to obtain a preset number of target eigenvectors.
Each target feature vector comprises a first feature vector and a second feature vector.
And a substep 208-13 of obtaining disaster matching scores corresponding to the detection points through a full connection layer included in the disaster prediction model based on the preset number of target feature vectors.
And a substep 208-14, determining a disaster prediction result of the remote sensing image to be analyzed according to the disaster matching score.
In addition, the detection indexes of the remote sensing image to be analyzed comprise: the disaster trend index of each known detection point in the remote sensing image to be analyzed and the disaster weight index in the remote sensing image to be analyzed are obtained; the detection indexes of the target disaster remote sensing image comprise: the disaster trend index of each detection point in the target disaster remote sensing image and the disaster weight index in the target disaster remote sensing image. The foregoing step 208 may also be implemented, for example, as follows to determine a disaster match score.
Determining a second mapping function of each detection point in the remote sensing image to be analyzed relative to each detection point in the target disaster remote sensing image based on the second matching confidence, wherein the second mapping function is used for reflecting the attention degree of each detection point in the remote sensing image to be analyzed to each detection point in the target disaster remote sensing image; determining candidate detection points from the target disaster remote sensing image according to the second mapping function and the disaster weight index of the detection points in the target disaster remote sensing image to obtain a second candidate detection point set; determining a second target detection point from the second candidate detection point set based on the known detection points in the remote sensing image to be analyzed; and generating corresponding detection points at corresponding positions in the remote sensing image to be analyzed based on the disaster trend index corresponding to the second target detection point, so as to predict the detection points of the remote sensing image to be analyzed and obtain a disaster prediction result.
On this basis, the foregoing sub-steps 208-13 can be realized by the following detailed description.
(1) And acquiring a target prediction vector through an attention mechanism included in the disaster prediction model based on the preset number of target feature vectors.
The target prediction vectors are determined according to a preset number of target feature vectors and a preset number of preset reference coefficients, and each target feature vector corresponds to one preset reference coefficient.
(2) And acquiring disaster matching scores corresponding to the detection points through a full-connection layer included in the disaster prediction model based on the target prediction vector.
In order to explain the present invention more clearly, for example, (1) in the above sub-steps 209 to 13 can be realized by the following embodiments.
Acquiring a preset number of first sub-feature vectors through a first sub-network included in an attention mechanism based on a preset number of target feature vectors, wherein the attention mechanism belongs to a disaster prediction model; acquiring a preset number of second sub-feature vectors through a second sub-network included in the attention mechanism based on the preset number of first sub-feature vectors; determining a preset number of preset reference coefficients according to a preset number of second sub-eigenvectors, wherein each preset reference coefficient corresponds to one target eigenvector; and determining a target prediction vector according to the preset number of target feature vectors and the preset number of reference coefficients.
In addition, the foregoing steps 208-12 may be implemented, for example, as follows.
Acquiring a preset number of first feature vectors through a first global average pooling layer included in a disaster prediction model based on the detection feature set, wherein each first feature vector corresponds to one detection feature; acquiring a preset number of second feature vectors through a second global average pooling layer included in the disaster prediction model based on the reference feature set, wherein each second feature vector corresponds to one reference feature; fusing a preset number of first eigenvectors and a preset number of second eigenvectors to obtain a preset number of target eigenvectors, wherein each target eigenvector comprises a first eigenvector and a second eigenvector; acquiring target prediction vectors through an attention mechanism included in a disaster prediction model based on a preset number of target feature vectors, wherein the target prediction vectors are determined according to the preset number of target feature vectors and a preset number of reference coefficients, and each target feature vector corresponds to one preset reference coefficient; and acquiring disaster matching scores corresponding to the detection points through a full-connection layer included in the disaster prediction model based on the target prediction vector.
As an alternative embodiment, the foregoing step 203 can be realized by the following specific steps.
And a substep 203-1 of constructing a logical relationship tree of known detection points in the original remote sensing image.
And a substep 203-2, determining the matching confidence degree between every two known detection points in the original remote sensing image according to the logical relationship tree of the known detection points.
And a substep 203-3 of determining a third mapping function of each known detection point in the original remote sensing image with respect to other known detection points based on the matching confidence between every two known detection points.
And the third mapping function is used for reflecting the attention degree of each detection point in the original remote sensing image to other known detection points in the track of the original remote sensing image.
And the substep 203-4 adjusts the logical relationship tree of the known detection points in the original remote sensing image according to a third mapping function to obtain the remote sensing image to be analyzed.
On the basis of the foregoing, in the foregoing step 204, determining the historical disaster remote sensing image based on the disaster data at the sample stage may be implemented by the following steps.
And a substep 204-1 of collecting disaster data at the sample stage.
And a substep 204-2, constructing a plurality of sample disaster remote sensing images according to the appointed acquisition stage and disaster data.
And a substep 204-3, integrating the plurality of sample disaster remote sensing images according to time, determining a detection point with the highest occurrence frequency at the same moment from the plurality of integrated sample disaster remote sensing images, and constructing and obtaining a target sample disaster remote sensing image according to the detection point with the highest occurrence frequency at the same moment.
And a substep 204-4, constructing a logical relationship tree of each detection point in the target sample disaster remote sensing image.
And a substep 204-5, determining the matching confidence coefficient between every two detection points in the target sample disaster remote sensing image according to the logical relationship tree of each detection point.
And a substep 204-6 of determining a fourth mapping function of each detection point in the target sample disaster remote sensing image relative to other detection points based on the matching confidence degrees between every two detection points.
And the fourth mapping function is used for reflecting the attention degree of each detection point in the target sample disaster remote sensing image to other detection points in the track of the target sample disaster remote sensing image.
And a substep 204-7, adjusting the logical relationship tree of the detection points in the target sample disaster remote sensing image according to a fourth mapping function to obtain a historical disaster remote sensing image.
In addition, for example, in the following embodiments, the step 204 can be realized by the following steps.
Acquiring disaster data of a sample stage; constructing a plurality of sample disaster remote sensing images according to the appointed acquisition stage and the disaster data; and processing the plurality of sample disaster remote sensing images through an attention mechanism to obtain historical disaster remote sensing images.
On the basis of the above, the computer device 100 is further connected to a disaster processing server in a communication manner, and in order to more clearly describe the solution of the present invention, the foregoing step 206 may be specifically implemented by the following steps.
And a substep 206-4 of obtaining the first target detection point, detection point information corresponding to the first target detection point, and a detection feature vector of the detection point information.
The detection feature vector of the detection point information includes first disaster data and at least one second disaster data of the first disaster data.
In sub-step 206-5, a data segment field configured in the inspection point information according to the first preset weight and the first preset disaster type is determined as the first disaster data.
In substep 206-6, at least one data segment field arranged along an edge of the first disaster data in the detected point information according to the second preset weight and the second preset disaster type is determined as second disaster data of the first disaster data.
Sub-step 206-7, finding a target disaster type matching the first target detection point.
Sub-step 206-8, creates a marking task for the first target detection point.
Wherein the marking task comprises an identification of the first target detection point.
Sub-step 206-9, pushing the marking task to the target disaster handling server.
Sub-step 206-10, receiving a marking request sent by the target disaster handling server to perform the marking task.
Wherein the marking request carries an identifier of the first target detection point.
And a substep 206-11 of returning the detection point information corresponding to the first target detection point and the detection feature vector of the detection point information to the target disaster processing server according to the identifier of the first target detection point, so that the target disaster processing server marks the first disaster data or the second disaster data of each detection feature vector of the detection point information to obtain a marked feature vector of the detection point information.
The mark feature vector of the detection point information comprises a mark object in the detection feature vector and a first sequence of the mark object in the detection feature vector.
And a sub-step 206-12 of receiving the marked feature vector of the detection point information submitted by the target disaster processing server.
And a substep 206-13 of converting the first sequence of the marked object in the detection feature vector to obtain a second sequence of the marked object in the first target detection point.
Sub-step 206-14, determining a marking field in the first target detection point according to the second sequence of marking objects in the first target detection point.
Sub-step 206-15, determining the second sequence of the marker object in the first target detection point and the marker field in the first target detection point as the marker feature vector of the first target detection point.
And a substep 206-16, determining a target disaster remote sensing image from the remote sensing image to be analyzed according to the marking characteristic vector of the detection point information and the marking characteristic vector of the first target detection point.
In addition to the foregoing aspects, embodiments of the present invention provide, for example, the following embodiments for determining a detected feature vector.
If a change request according to the first disaster data sent by the target disaster processing server is received, acquiring the changed first disaster data, and re-determining at least one second disaster data of the changed first disaster data; and updating the detection feature vector of the detection point information according to the changed first disaster data and at least one second disaster data of the changed first disaster data.
In order to improve the efficiency of data processing, the following embodiments are provided.
Performing correlation storage on the first target detection point and the marking feature vector of the first target detection point; and inputting the first target detection point and the mark characteristic vector of the first target detection point which are stored in an associated manner into a machine learning model as sample data to perform machine learning.
An embodiment of the present invention provides a data processing system 110 for remote sensing monitoring of natural disasters, which is applied to a computer device 100, wherein the computer device 100 is in communication connection with a remote sensing device 200, as shown in fig. 3, the data processing system 110 for remote sensing monitoring of natural disasters includes:
an obtaining module 1101, configured to obtain a known detection point collected by the remote sensing device 200 in a current collection phase; determining an original remote sensing image based on known detection points in the current acquisition stage; and carrying out remote sensing information processing on the original remote sensing image through a preset neural network algorithm to obtain a remote sensing image to be analyzed.
The determining module 1102 is configured to determine a first mapping matrix between detection indexes of a remote sensing image to be analyzed and detection indexes of a remote sensing image of a historical disaster, and includes: respectively constructing a logical relationship tree of each detection point in a remote sensing image to be analyzed and a historical disaster remote sensing image to obtain a plurality of first remote sensing characteristic locus diagrams and a plurality of second remote sensing characteristic locus diagrams, determining a first matching confidence coefficient between each first remote sensing characteristic locus diagram and each second remote sensing characteristic locus diagram to obtain a first mapping matrix, wherein the historical disaster remote sensing image is determined based on disaster data at a sample stage; extracting candidate detection points meeting preset association rules of the remote sensing image to be analyzed from the historical disaster remote sensing image to obtain a first candidate detection point set; determining a first target detection point from the first candidate detection point set based on the first mapping matrix, and adjusting detection point information of the remote sensing image to be analyzed according to the first target detection point to obtain a target disaster remote sensing image; determining a second mapping matrix between the detection index of the target disaster remote sensing image and the detection index of the remote sensing image to be analyzed, wherein the second mapping matrix comprises the following steps: and respectively constructing a logical relationship tree of each detection point in the remote sensing image to be analyzed and the target disaster remote sensing image to obtain a plurality of third remote sensing characteristic locus diagrams and a plurality of fourth remote sensing characteristic locus diagrams, determining a second matching confidence coefficient between each third remote sensing characteristic locus diagram and each fourth remote sensing characteristic locus diagram, and obtaining a second mapping matrix.
And the prediction module 1103 is configured to perform detection point prediction on the remote sensing image to be analyzed according to the second mapping matrix and the detection points in the remote sensing image to be analyzed, so as to obtain a disaster prediction result.
Further, the detection indexes of the remote sensing image to be analyzed comprise: the disaster trend index of each known detection point in the remote sensing image to be analyzed and the disaster weight index in the remote sensing image to be analyzed are obtained; the detection indexes of the remote sensing image of the historical disaster comprise: the disaster trend index of each detection point in the historical disaster remote sensing image and the disaster weight index in the historical disaster remote sensing image.
The determining module 1102 is specifically configured to:
determining a first mapping function of each detection point in the remote sensing image to be analyzed relative to each detection point in the historical disaster remote sensing image based on the first matching confidence, wherein the first mapping function is used for reflecting the attention degree of each detection point in the remote sensing image to be analyzed to each detection point in the historical disaster remote sensing image; determining corresponding candidate detection points from the first candidate detection point set according to the sequence of the first mapping function from high to low, wherein the corresponding candidate detection points serve as first target detection points; and generating corresponding detection points at corresponding positions in the remote sensing image to be analyzed according to the disaster weight index of the first target detection points in the historical disaster remote sensing image and the disaster trend index corresponding to the first target points.
Further, the prediction module 1103 is specifically configured to:
acquiring detection points, and acquiring corresponding second mapping matrixes according to the detection points; based on detection points, obtaining a detection feature set through a first feature extraction network included in a disaster prediction model, wherein the detection feature set comprises a preset number of detection features; acquiring a reference feature set through a second feature extraction network included in the disaster prediction model based on a second mapping matrix, wherein the reference feature set includes a preset number of reference features; based on the detection feature set, according to each frame of detection features in the detection feature set, obtaining a first maximum pooling feature value through a maximum pooling layer included in a first attention mechanism, wherein the first attention mechanism belongs to a disaster prediction model; according to each frame of detection features in the detection feature set, a first average pooling feature value is obtained through a global average pooling layer included in a first attention mechanism; acquiring a first fusion feature through a convolutional layer included in a first attention mechanism based on a first maximum pooling feature value and a first average pooling feature value according to each frame detection feature in a detection feature set; according to each frame of detection features in the detection feature set, based on the first fusion features and the detection features, obtaining first feature vectors through a first global average pooling layer included in a first attention mechanism, wherein each first feature vector corresponds to one detection feature; acquiring a second maximum pooling characteristic value through a maximum pooling layer included in a second attention mechanism according to each frame of reference feature in the reference feature set, wherein the second attention mechanism belongs to a disaster prediction model; acquiring a second average pooling characteristic value through a global average pooling layer included in a second attention mechanism according to each frame of reference characteristic in the reference characteristic set; acquiring a second fusion feature through a convolutional layer included in a second attention mechanism based on a second maximum pooling feature value and a second average pooling feature value according to each frame of reference feature in the reference feature set; acquiring second feature vectors through a second global average pooling layer included in a second attention mechanism according to each frame of reference feature in the reference feature set based on the second fusion feature and the reference feature, wherein each second feature vector corresponds to one reference feature; fusing a preset number of first eigenvectors and a preset number of second eigenvectors to obtain a preset number of target eigenvectors, wherein each target eigenvector comprises a first eigenvector and a second eigenvector; acquiring disaster matching scores corresponding to detection points through a full-connection layer included in a disaster prediction model based on a preset number of target feature vectors; and determining a disaster prediction result of the remote sensing image to be analyzed according to the disaster matching score.
Further, the prediction module 1103 is further configured to:
acquiring target prediction vectors through an attention mechanism included in a disaster prediction model based on a preset number of target feature vectors, wherein the target prediction vectors are determined according to the preset number of target feature vectors and a preset number of reference coefficients, and each target feature vector corresponds to one preset reference coefficient; and acquiring disaster matching scores corresponding to the detection points through a full-connection layer included in the disaster prediction model based on the target prediction vector.
Further, the obtaining module 1101 is specifically configured to:
constructing a logic relation tree of known detection points in an original remote sensing image; determining matching confidence between every two known detection points in the original remote sensing image according to the logical relationship tree of the known detection points; determining a third mapping function of each known detection point in the original remote sensing image relative to other known detection points based on the matching confidence degree between every two known detection points, wherein the third mapping function is used for reflecting the attention degree of each detection point in the original remote sensing image to other known detection points in the track of the detection point; and adjusting the logical relation tree of the known detection points in the original remote sensing image according to the third mapping function to obtain the remote sensing image to be analyzed.
Further, the historical disaster prediction module 1103, which is determined based on the disaster data at the sample stage, is specifically configured to:
disaster data of a sample stage are collected; constructing a plurality of sample disaster remote sensing images according to the appointed acquisition stage and disaster data; integrating the sample disaster remote sensing images according to time, determining a detection point with highest occurrence frequency at the same moment from the integrated sample disaster remote sensing images, and constructing and obtaining a target sample disaster remote sensing image according to the detection point with the highest occurrence frequency at the same moment; constructing a logical relation tree of each detection point in the target sample disaster remote sensing image; determining a matching confidence coefficient between every two detection points in the target sample disaster remote sensing image according to the logic relation tree of each detection point; determining a fourth mapping function of each detection point in the target sample disaster remote sensing image relative to other detection points based on the matching confidence coefficient between every two detection points, wherein the fourth mapping function is used for reflecting the attention degree of each detection point in the target sample disaster remote sensing image to other detection points in the track of the detection point; and adjusting the logical relation tree of the detection points in the target sample disaster remote sensing image according to the fourth mapping function to obtain the historical disaster remote sensing image.
Further, the computer device 100 is further communicatively connected to a disaster processing server, and the determining module 1102 is specifically configured to:
acquiring a first target detection point, detection point information corresponding to the first target detection point and a detection feature vector of the detection point information, wherein the detection feature vector of the detection point information comprises first disaster data and at least one second disaster data of the first disaster data; determining a data segment field configured in the detection point information according to the first preset weight and the first preset disaster type as first disaster data; determining at least one data segment field configured along the edge of the first disaster data in the detection point information according to a second preset weight and a second preset disaster type as second disaster data of the first disaster data; searching a target disaster type matched with the first target detection point; creating a marking task for the first target detection point, wherein the marking task comprises an identifier of the first target detection point; pushing the marking task to a target disaster processing server; receiving a marking request sent by a target disaster processing server for executing a marking task, wherein the marking request carries an identifier of a first target detection point; returning the detection point information corresponding to the first target detection point and the detection feature vector of the detection point information to the target disaster processing server according to the identifier of the first target detection point, so that the target disaster processing server marks the first disaster data or the second disaster data of each detection feature vector of the detection point information to obtain a marked feature vector of the detection point information, wherein the marked feature vector of the detection point information comprises a marked object in the detection feature vector and a first sequence of the marked object in the detection feature vector; receiving a marked feature vector of detection point information submitted by a target disaster processing server; converting a first sequence of the marked object in the detection feature vector to obtain a second sequence of the marked object in the first target detection point; determining a marking field in the first target detection point according to the second sequence of the marking object in the first target detection point; determining a second sequence of the mark object in the first target detection point and a mark field in the first target detection point as a mark feature vector of the first target detection point; and determining a target disaster remote sensing image from the remote sensing image to be analyzed according to the marked feature vector of the detection point information and the marked feature vector of the first target detection point.
It should be noted that, as for the foregoing implementation principle of the data processing system 110 for remote sensing monitoring natural disasters, reference may be made to the implementation principle of the data processing method for remote sensing monitoring natural disasters, which is not described herein again. It should be understood that the division of the modules of the above system is only a logical division, and the actual implementation may be wholly or partially integrated into one physical entity or may be physically separated. And these modules can be realized in the form of software called by processing element; or may be implemented entirely in hardware; and part of the modules can be realized in the form of calling software by the processing element, and part of the modules can be realized in the form of hardware. For example, the obtaining module 1101 may be a processing element separately set up, or may be implemented by being integrated into a chip of the system, or may be stored in a memory of the system in the form of program code, and the function of the obtaining module 1101 may be called and executed by a processing element of the system. Other modules are implemented similarly. In addition, all or part of the modules can be integrated together or can be independently realized. The processing element described herein may be an integrated circuit having signal processing capabilities. In implementation, each step of the above method or each module above may be implemented by an integrated logic circuit of hardware in a processor element or an instruction in the form of software.
For example, the above modules may be one or more integrated circuits configured to implement the above methods, such as: one or more Application Specific Integrated Circuits (ASICs), or one or more microprocessors (DSPs), or one or more Field Programmable Gate Arrays (FPGAs), among others. For another example, when some of the above modules are implemented in the form of a processing element scheduler code, the processing element may be a general-purpose processor, such as a Central Processing Unit (CPU) or other processor that can call program code. As another example, these modules may be integrated together, implemented in the form of a system-on-a-chip (SOC).
The embodiment of the present invention provides a computer device 100, where the computer device 100 includes a processor and a non-volatile memory storing computer instructions, and when the computer instructions are executed by the processor, the computer device 100 executes the remote sensing monitoring natural disaster data processing system 110. As shown in fig. 4, fig. 4 is a block diagram of a computer device 100 according to an embodiment of the present invention. The computer device 100 comprises a data processing system 110 for remotely monitoring natural disasters, a memory 111, a processor 112 and a communication unit 113.
To facilitate the transfer or interaction of data, the elements of the memory 111, the processor 112 and the communication unit 113 are electrically connected to each other, directly or indirectly. For example, these components may be electrically connected to each other via one or more communication buses or signal lines. The data processing system 110 for remote monitoring natural disasters includes at least one software function module which can be stored in a memory 111 in the form of software or firmware (firmware) or solidified in an Operating System (OS) of the computer device 100. The processor 112 is configured to execute the data processing system 110 for remote sensing natural disaster monitoring stored in the memory 111, for example, software function modules and computer programs included in the data processing for remote sensing natural disaster monitoring.
The readable storage medium comprises a computer program, and when the computer program runs, the computer device 100 where the readable storage medium is located is controlled to execute the method for processing the remote sensing monitoring natural disaster data.
In summary, the invention discloses a method and a system for processing data of remote sensing monitoring natural disasters, which are implemented by acquiring known detection points acquired by remote sensing equipment 200 in a current acquisition stage; further determining an original remote sensing image and a remote sensing image to be analyzed; by determining the remote sensing image to be analyzed and the historical disaster remote sensing image, obtaining the target disaster remote sensing image, and determining the second mapping matrix between the detection index of the target disaster remote sensing image and the detection index of the remote sensing image to be analyzed, the detection point prediction can be carried out on the remote sensing image to be analyzed according to the second mapping matrix and the detection point in the remote sensing image to be analyzed, so that the disaster prediction result obtained through prediction can be determined based on the remote sensing image to be analyzed and the historical disaster remote sensing image, and the disaster prediction result with more credible reference degree can be obtained.
The foregoing description, for purpose of explanation, has been described with reference to specific embodiments. However, the illustrative discussions above are not intended to be exhaustive or to limit the disclosure to the precise forms disclosed. Many modifications and variations are possible in light of the above teaching. The embodiments were chosen and described in order to best explain the principles of the disclosure and its practical applications, to thereby enable others skilled in the art to best utilize the disclosure and various embodiments with various modifications as are suited to the particular use contemplated. The foregoing description, for purpose of explanation, has been described with reference to specific embodiments. However, the illustrative discussions above are not intended to be exhaustive or to limit the disclosure to the precise forms disclosed. Many modifications and variations are possible in light of the above teaching. The embodiments were chosen and described in order to best explain the principles of the disclosure and its practical applications, to thereby enable others skilled in the art to best utilize the disclosure and various embodiments with various modifications as are suited to the particular use contemplated.