An image data processing method based on big data and artificial intelligence generation system technology

Abstract

The present invention relates to the technical field of remote sensing image processing, and discloses an image data processing method based on big data and artificial intelligence generation system technology, comprising the following steps: Step 100, collecting motion estimation information and glacier animal habitat information; Step 200, encoding the motion estimation information into first graph structure data; Step 300, inputting the first graph structure data into a motion estimation model, where the first output layer outputs a result representing the classification of the motion pattern of the glacier, and the second output layer outputs a result representing the movement of the surface features of the glacier. The present invention comprehensively analyzes the results of the glacier motion pattern and the movement of the surface features, and can more comprehensively understand the dynamic characteristics, mass balance state, response to environmental changes, and impact on regional water resources of the glacier.

Description

Technical Field

[0001] The present invention relates to the field of remote sensing image processing, and more specifically, to an image data processing method based on big data and artificial intelligence generation system technology. Background Art

[0002] A glacier refers to a natural ice body that exists on the surface of polar or alpine regions for many years and has a state of movement along the ground. Glaciers are formed by years of accumulated snow, which undergoes processes such as compaction, recrystallization, and refreezing. It has a certain shape and layer structure, and is plastic. Under the action of gravity and pressure, it undergoes plastic flow and block sliding, and is an important fresh water resource on the earth's surface.

[0003] Especially, glaciers in polar regions have complex non-linear motion behaviors such as glacier surges and tidal effects, which are difficult to describe with simple models. At the same time, the motion in areas such as glacier edges and cracks is discontinuous, resulting in instability in motion field estimation. Moreover, there is large-scale three-dimensional flow in some areas such as the bottom of ice shelves, and simple surface motion analysis may not be comprehensive enough, causing difficulties in glacier research, disaster risk assessment, and water resource management research. Summary of the Invention

[0004] The present invention provides an image data processing method based on big data and artificial intelligence generation system technology to solve the technical problems in related technologies.

[0005] The present invention provides an image data processing method based on big data and artificial intelligence generation system technology, including the following steps:

[0006] Step 100, collecting motion estimation information and glacier animal habitat information; the motion estimation information includes optical and SAR satellite image information, satellite image imaging information, glacier information, and bed information; the glacier animal habitat information includes animal species, animal habitat locations, animal habitat areas, and animal numbers;

[0007] Step 200, encoding the motion estimation information into first graph structure data, the first graph structure data includes nodes and edges connecting the nodes, and a node represents a satellite image, a glacier, a bed, or a satellite;

[0008] Constructing second graph structure data, the second graph structure data includes nodes and edges connecting the nodes, and a node represents an animal or a habitat; encoding the glacier animal habitat information of each node into sequence data respectively, the sequence data includes n sequence units, and the t-th sequence unit represents the glacier animal habitat information of a node at the t-th moment;

[0009] Step 300: Input the first map structure data into the motion estimation model. The motion estimation model includes a first hidden layer, a first output layer, and a second output layer. The first output layer outputs the result representing the classification of the motion pattern of the glacier, and the second output layer outputs the result representing the movement of the surface features of the glacier.

[0010] Input the sequence data, the output result of the first output layer, and the output result of the second output layer into the input habitat assessment model. The third output layer outputs the change rate of the core area of the species.

[0011] Step 400: Based on the output result of the classification of the motion pattern and the result of the movement of the surface features of the glacier, evaluate the sensitivity of the glacier to environmental changes, the stability of the glacier, and the potential risks, or the impact of glacier changes on regional water resources.

[0012] Furthermore, the satellite image imaging information includes: satellite orbit parameters, imaging geometric information, and ground control point information of the imaging satellite.

[0013] Glacier information: glacier boundary, glacier surface elevation, glacier surface velocity, glacier thickness, glacier mass balance, glacier surface temperature, glacier surface albedo, and glacier surface morphology.

[0014] Bedrock information: bedrock topography, bedrock elevation, bedrock slope, bedrock roughness, bedrock type, bedrock temperature, and bedrock stress.

[0015] Furthermore, there is an edge between the node representing an animal and all the nodes representing habitats. There is an edge between all the nodes corresponding to the animals in the same node representing a habitat. There is an edge between all the nodes representing habitats.

[0016] Furthermore, there is an edge between the node representing the glacier in the same satellite image and the node representing the bedrock.

[0017] There is an edge between the node representing the glacier in the satellite image and the node representing the glacier in the satellite image of the previous timestamp.

[0018] If the similarity between the node representing the glacier in the satellite image and the node representing the glacier in the satellite image of the previous timestamp exceeds the set threshold, it is determined that there is an edge between them.

[0019] If the similarity between the nodes representing the bedrock in the satellite image exceeds the set threshold, it is determined that there is an edge between them.

[0020] There is an edge between the satellite image and the satellite.

[0021] Furthermore, there are edges between the nodes representing animals and all the nodes representing habitats, between the nodes corresponding to all the animals in the same node representing a habitat, and between all the nodes representing habitats.

[0022] The habitat assessment model runs twice. When running for the first time, the first feature fusion layer inputs the second hidden state and the default feature, and the size of the default feature is the same as that of the second hidden state. The first output layer outputs the initial change rate of the core area of the species.

[0023] When running for the second time, the first feature fusion layer inputs the second hidden state, the fourth hidden state, and the fifth hidden state, and the third output layer outputs the change rate of the core area of the species for the current situation.

[0024] Furthermore, the calculation formula of the first hidden layer is as follows:

[0025] ;

[0026] where represents the first hidden state of the v-th node, represents the graph weight parameter in the first hidden layer, and respectively represent the node features of the v-th and u-th nodes of the first graph structure data, is the set of nodes that have edges with the node v of the first graph structure data, represents the total number of nodes in;

[0027] ;

[0028] where represents the aggregation coefficient between the v-th and u-th nodes, tanh represents the hyperbolic tangent function, T represents the transpose, represents the adjustable parameter; The default value of is 1 / N, where N is the dimension of the node feature;

[0029] The calculation formula of the first output layer is as follows:

[0030] ;

[0031] where represents the first output vector, and the i-th component of the first output vector represents the probability that the movement mode of the glacier belongs to the i-th state, represents the v-th first hidden state input to the first output layer, represents the concatenation of the first hidden states of all the nodes input to the first output layer, represents the set of all nodes, is the weight parameter in the first output layer, is the bias parameter in the first output layer, represents the sigmoid function;

[0032] The calculation formula of the second output layer is as follows:

[0033] ;

[0034] where represents the second output vector, and the i-th component of the second output vector represents the probability value of the i-th action; the set of all actions is called the action space, which contains all possible actions that can be selected;

[0035] In an embodiment of the present invention, an action is represented as a vector, and the j-th component of the vector represents the j-th representation parameter of the movement of the surface feature of the glacier;

[0036] An action is represented as a matrix. The element in the first row and the L-th column of the matrix represents the representation parameter of the L-th displacement vector, the element in the second row and the L-th column represents the representation parameter of the L-th velocity field, the element in the third row and the L-th column represents the representation parameter of the L-th strain rate, the element in the fourth row and the L-th column represents the representation parameter of the L-th surface elevation change, and the element in the fifth row and the L-th column represents the representation parameter of the L-th surface morphology change. represents the first hidden state of the v-th node input to the second output layer, represents the concatenation of the first hidden states of all nodes input to the second output layer, and Z represents the set of nodes related to the recognition result, is the weight parameter in the second output layer, is the bias parameter in the second output layer, represents the sigmoid function.

[0037] Furthermore, the calculation formula of the second hidden layer is as follows:

[0038] ;

[0039] where, represents the second graph structure data recognition feature of the d-th node in the l-th layer of the t-th sequence unit of the sequence data, represents the second graph structure data recognition feature of the u-th node in the t-th sequence unit of the sequence data in the l-1-th layer, and respectively represent the sets of nodes with high association relationships with node d and node u, represents the cardinality of the set, represents the weight matrix of the second graph structure data in the l-th layer, , E represents the total number of layers, when l = 1 , The feature representation of an object indicating the u-th node index of the t-th sequence unit of sequence data, is the sigmoid function;

[0040] The calculation formula for the fourth hidden layer is as follows:

[0041] ;

[0042] ;

[0043] where, represents the fourth hidden state, represents the information of the movement pattern of the input glacier, represents the first convolution function;

[0044] The calculation formula for the fifth hidden layer is as follows:

[0045] ;

[0046] ;

[0047] where, represents the fifth hidden state, represents the information of the movement of the surface features of the input glacier, represents the second convolution function.

[0048] Furthermore, the calculation formula for the first feature fusion layer is as follows:

[0049] ;

[0050] where, represents the first fusion state, represents the recognition feature of the last layer of the d-th node output when the t-th sequence unit of the sequence data is input to the second hidden layer, represents the set of all units of the second graph structure data, represents the concatenation function, represents the summation weight matrix, represents the summation bias parameter;

[0051] The calculation formula for the fifth hidden layer is as follows:

[0052] ;

[0053] ;

[0054] ;

[0055] ;

[0056] Among them, , , represent the first, second, and third weight matrices in the fifth hidden layer, , , represent the first, second, and third bias parameters in the fifth hidden layer, represents the dot product, , and respectively represent the first, second, and third intermediate states in the fifth hidden layer, , and respectively represent the t-th and (t - 1)-th fifth hidden states, where n ≥ t ≥ 1, n represents the total number of sequence units of the sequence data, and when t = 1 , tanh is the hyperbolic tangent function, represents the sigmoid function;

[0057] The calculation formula of the third output layer is as follows:

[0058] ;

[0059] Among them represents the output matrix, and an element of the output matrix represents the change rate of the species core area of the species near a glacier, represents the g-th fifth hidden state, represents the concatenation of the fifth hidden states of all nodes, represents the set of all nodes, is the weight parameter in the third output layer, is the bias parameter in the third output layer, represents the S function.

[0060] The present invention also proposes an image data processing system based on the generation system technology of big data and artificial intelligence, which executes the steps in the foregoing image data processing method based on the generation system technology of big data and artificial intelligence, including:

[0061] Data acquisition module: acquiring motion estimation information and glacier animal habitat information based on the generation system technology of big data and artificial intelligence;

[0062] Data encoding module: constructing the first graph structure data, sequence data, and second graph structure data, including sequences or nodes and edges, as the input for establishing corresponding motion estimation models and habitat evaluation models for each glacier and bedrock and habitats and animals;

[0063] Data processing module: Input the encoded first graph structure data, sequence data, and second graph structure data into the motion estimation model and the habitat assessment model, and output the corresponding results of motion pattern classification and the results of the movement of the surface features of the glacier.

[0064] Result evaluation module: Evaluate the dynamic characteristics, mass balance state, response to environmental changes, and impact on regional water resources of the glacier through the corresponding results of motion pattern classification and the results of the movement of the surface features of the glacier. At the same time, it can also combine glacier animal habitat information to obtain the change rate of the core area of the species, and evaluate the impact of glacier changes on the species near the glacier.

[0065] A storage medium stores non-temporary computer-readable instructions for performing the steps in the aforementioned image data processing method based on big data and artificial intelligence generation system technology.

[0066] The beneficial effects of the present invention are as follows:

[0067] The present invention comprehensively analyzes the results of glacier motion patterns and surface feature movements, and can more comprehensively understand the dynamic characteristics, mass balance state, response to environmental changes, and impact on regional water resources of the glacier. Brief Description of the Drawings

[0068] Figure 1 is a flowchart of an image data processing method based on big data and artificial intelligence generation system technology proposed by the present invention.

[0069] Figure 2 is a structural block diagram of an image data processing system based on big data and artificial intelligence generation system technology proposed by the present invention.

[0070] In the figure: 101, data acquisition module; 102, data encoding module; 103, data processing module; 104, result evaluation module. Detailed Embodiments

[0071] Now, the subject matter described herein will be discussed with reference to exemplary embodiments. It should be understood that discussing these embodiments is only to enable those skilled in the art to better understand and thus implement the subject matter described herein. Without departing from the scope of protection of the content of this specification, changes can be made to the functions and arrangements of the elements discussed. Each example can omit, substitute, or add various processes or components as needed. Additionally, the features described relative to some examples can also be combined in other examples.

[0072] As Figure 1 shown, an image data processing method based on big data and artificial intelligence generation system technology includes.

[0073] Step 100, collect motion estimation information and glacier animal habitat information; the motion estimation information includes optical and SAR satellite image information, satellite image imaging information, glacier information and bedrock information; the glacier animal habitat information includes animal species, animal habitat locations, animal habitat areas and animal numbers;

[0074] The satellite image imaging information includes: satellite orbit parameters, imaging geometric information, and ground control point information of the imaging satellite;

[0075] Glacier information: glacier boundary, glacier surface elevation, glacier surface velocity, glacier thickness, glacier mass balance, glacier surface temperature, glacier surface albedo and glacier surface morphology;

[0076] Bedrock information: bedrock topography, bedrock elevation, bedrock slope, bedrock roughness, bedrock type, bedrock temperature and bedrock stress;

[0077] Step 200, encode the motion estimation information into first graph structure data. The first graph structure data includes nodes and edges connecting the nodes. A node represents a satellite image, a glacier, a bedrock or a satellite;

[0078] There is an edge between the node representing the glacier in the same satellite image and the node representing the bedrock;

[0079] There is an edge between the node representing the glacier in the satellite image and the node representing the glacier in the satellite image of the previous timestamp;

[0080] If the similarity between the node representing the glacier in the satellite image and the node representing the glacier in the satellite image of the previous timestamp exceeds a set threshold, it is determined that there is an edge between them;

[0081] If the similarity between the nodes representing the bedrock in the satellite image exceeds a set threshold, it is determined that there is an edge between them;

[0082] There is an edge between the satellite image and the satellite;

[0083] Construct second graph structure data. The second graph structure data includes nodes and edges connecting the nodes. A node represents an animal or a habitat;

[0084] There is an edge between the node representing an animal and all the nodes representing habitats, there is an edge between the nodes corresponding to all the animals connecting to the same node representing a habitat, and there is an edge between all the nodes representing habitats;

[0085] Encode the glacier animal habitat information of each node into sequence data respectively. The sequence data includes n sequence units, and the t-th sequence unit represents the glacier animal habitat information of a node at the t-th moment;

[0086] Step 300: Input the first image structure data into a motion estimation model. The motion estimation model includes a first hidden layer, a first output layer, and a second output layer. The first output layer outputs a result representing the classification of the motion pattern of the glacier, and the second output layer outputs a result representing the movement of the surface features of the glacier.

[0087] In an embodiment of the present invention, the classification results of the motion pattern of the glacier include block sliding, internal deformation, creep, and periodic surging.

[0088] It should be noted that block sliding means that the bottom of the glacier slides along the surface of the bed under the action of gravity, which is the main motion mode of valley glaciers.

[0089] Internal deformation means that plastic deformation occurs inside the glacier, and the upper ice body moves relative to the lower ice body, causing deformation of the entire glacier.

[0090] Creep means that under the action of the self-weight stress of the glacier, the ice body undergoes slow creep deformation.

[0091] Periodic surging means that some glaciers exhibit a phenomenon of alternating periodic rapid advance and relative stillness.

[0092] In an embodiment of the present invention, the result of the movement of the surface features of the glacier can be measured by one or a combination of displacement vectors, velocity fields, strain rates, surface elevation changes, and surface morphology changes. The result can be expressed as a combination of one or more parameters such as the overall movement direction and distance of the glacier, the movement speed and direction of the glacier surface, the degree and direction of glacier deformation, the change of the surface morphological features of the glacier, and the change of the surface height or volume of the glacier.

[0093] It should be noted that the displacement vector uses a vector to represent the movement direction and distance of the surface features of the glacier (such as cracks, glacier wells, etc.).

[0094] The starting point of the vector represents the initial position of the feature, and the end point represents the final position of the feature.

[0095] The length of the vector represents the movement distance of the feature, and the direction represents the movement direction.

[0096] Among them, the velocity field uses a velocity vector field to represent the movement speed and direction of different positions on the glacier surface.

[0097] Each grid point or pixel has a velocity vector, representing the movement speed and direction of the glacier surface at that point.

[0098] The velocity field can be used to analyze the spatial pattern and temporal variation of the movement of the glacier surface.

[0099] Among them, the strain rate is calculated for different regions on the glacier surface, representing the degree and direction of glacier deformation.

[0100] The strain rate can be decomposed into a transverse strain rate (indicating the transverse stretching or compression of the glacier) and a longitudinal strain rate (indicating the longitudinal stretching or compression of the glacier);

[0101] A positive value of the strain rate indicates stretching, and a negative value indicates compression.

[0102] Among them, the surface elevation change is calculated by repeatedly measuring the glacier surface elevation to obtain the elevation change between different periods.

[0103] The elevation change can be expressed as the vertical displacement or thickness change of the glacier surface;

[0104] A positive value indicates the glacier surface rising or thickening, and a negative value indicates the glacier surface sinking or thinning.

[0105] Among them, the surface morphology change is to analyze the changes in the surface morphological characteristics of the glacier (such as glacier cracks, glacier wells, glacier springs, etc.).

[0106] By calculating the changes in indicators such as the quantity, density, and direction of the glacier surface morphological characteristics at different times, the surface morphology change can reflect the changes in the glacier dynamics process and environmental conditions.

[0107] Among them, the mass balance is calculated for the mass balance of the glacier surface, that is, the difference between the accumulation amount (such as snowfall) and the ablation amount (such as melting, sublimation).

[0108] The mass balance can be expressed as the change in the glacier surface height or volume;

[0109] A positive value indicates an increase in the glacier mass, and a negative value indicates a loss in the glacier mass.

[0110] In an example of the present invention, it is necessary to study a glacier located in Antarctica. Through GPS measurements and remote sensing image analysis of the glacier surface, the following results of the movement of the glacier surface characteristics are obtained:

[0111] Displacement vector: By tracking the large cracks on the glacier surface, it is found that the cracks move about 200 meters downstream within one year; the direction of the displacement vector points to the downstream direction of the glacier, indicating that the glacier as a whole flows downstream;

[0112] Velocity field: By repeatedly measuring the characteristic points on the glacier surface, the velocity field of the glacier surface is calculated; the velocity field shows that the velocity at the center line of the glacier is the largest, reaching 300 meters per year, while the velocity near the glacier edge is slower, about 50 meters per year; the spatial pattern of the velocity field indicates that the flow of the glacier is mainly affected by the terrain and the glacier bed.

[0113] Strain rate: Calculate the strain rate of the glacier surface based on the velocity field; the results show that the transverse strain rate in the upper part of the glacier is positive, indicating that the glacier is being stretched laterally; the longitudinal strain rate in the lower part of the glacier is negative, indicating that the glacier is being compressed longitudinally.

[0114] Surface elevation change: Obtain the change in the glacier surface elevation through repeated airborne lidar measurements; the results show that the surface elevation in the upper part of the glacier has decreased by 2 meters within a year, indicating that the glacier is thinning; the change in the surface elevation in the lower part of the glacier is relatively small, indicating that the glacier thickness is relatively stable.

[0115] Surface morphology change: Through the analysis of high-resolution satellite images, it is found that the number and density of cracks on the glacier surface have increased within a year; new cracks mainly appear in the upper part of the glacier, indicating that the glacier movement in this area is more active; at the same time, it is observed that the number of glacier wells on the glacier surface has decreased, indicating an increase in the ablation intensity of the glacier.

[0116] Through the above parameter combinations, comprehensively evaluate the results of the movement of glacier surface features, and infer the dynamic process of the glacier and its response to environmental changes;

[0117] Among them, the first hidden layer inputs the first graph structure data and outputs the first hidden state to the first output layer. The first output layer outputs the classification result of the glacier movement mode. The first hidden state is input to the second output layer, and the second output layer outputs the result of the movement of glacier surface features.

[0118] Input the sequence data, the output result of the first output layer, and the output result of the second output layer into the input habitat assessment model. The habitat assessment model includes a second hidden layer, a third hidden layer, a fourth hidden layer, a fifth hidden layer, a first feature fusion layer, and a third output layer. The third output layer outputs the change rate of the species core area.

[0119] It should be added that the change rate of the species core area is the ratio of the difference between the current core area and the past core area to the past core area;

[0120] Among them, the second hidden layer inputs the sequence data and outputs the second hidden state. The output result of the first output layer and the output result of the second output layer are respectively input to the fourth hidden layer and the fifth hidden layer. Then, the fourth hidden state and the fifth hidden state output by the fourth hidden layer and the fifth hidden layer are input to the first feature fusion layer together with the second hidden state. The first feature fusion layer outputs the first fusion state. The output first fusion state is input to the third hidden layer. The third hidden layer outputs the third hidden state to the third output layer. The third output layer outputs the result representing the change rate of the five-core area;

[0121] The habitat assessment model runs twice. When running for the first time, the first feature fusion layer inputs the second hidden state and the default feature, and the size of the default feature is the same as that of the second hidden state. The first output layer outputs the initial change rate of the core area of the species.

[0122] When running for the second time, the first feature fusion layer inputs the second hidden state, the fourth hidden state, and the fifth hidden state, and the third output layer outputs the change rate of the core area of the species for the current situation.

[0123] Step 400: According to the results of the classified motion patterns and the results of the movement of the glacier's surface features, evaluate the sensitivity of the glacier to environmental changes, the stability of the glacier, and potential risks, or the impact of glacier changes on regional water resources.

[0124] Specifically, the sensitivity of the glacier to environmental changes can be evaluated by comprehensively analyzing the changes in glacier movement and surface features, and the sensitivity of the glacier to environmental changes (such as climate warming) can be assessed.

[0125] Accelerated glacier flow, increased strain rate, intensified surface ablation, etc. may all indicate an enhanced response of the glacier to environmental changes.

[0126] The stability and potential risks of the glacier are to evaluate the overall stability of the glacier, which is very important for predicting glacier-related disasters (such as ice avalanches, glacier floods, etc.); accelerated glacier flow, increased crack density, unstable mass balance, etc. may all indicate a decrease in glacier stability and an increase in potential risks.

[0127] The impact of glacier changes on regional water resources is that glacier changes will affect the regional water resource balance and hydrological processes; accelerated glacier ablation may lead to an increase in river runoff in the short term, but may lead to a decrease in water resources in the long term; evaluating the impact of glacier changes on water resources is crucial for regional water resource management and planning.

[0128] By comprehensively analyzing the results of glacier movement patterns and surface feature movements, a more comprehensive understanding of the dynamic characteristics, mass balance state, response to environmental changes, and impact on regional water resources of the glacier can be obtained, which is of great significance for fields such as glacier research, disaster risk assessment, and water resource management.

[0129] Based on the above results, the research on species in the habitats near the glacier can also be output using the habitat assessment model.

[0130] In an embodiment of the present invention, the calculation formula of the first hidden layer is as follows:

[0131] ;

[0132] Where represents the first hidden state of the v-th node, represents the graph weight parameter in the first hidden layer, and respectively represent the node features of the v-th and u-th nodes of the first graph structure data, is the set of nodes that have edges with node v of the first graph structure data, represents the total number of nodes in

[0133] ;

[0134] where represents the aggregation coefficient between the v-th and u-th nodes, tanh represents the hyperbolic tangent function, T represents the transpose, represents the adjustable parameter; The default value of

[0135] In one embodiment of the present invention, the calculation formula of the first output layer is as follows:

[0136] ;

[0137] where represents the first output vector, and the i-th component of the first output vector represents the probability that the movement pattern of the glacier belongs to the i-th state, represents the v-th first hidden state input to the first output layer, represents concatenating the first hidden states of all nodes input to the first output layer, represents the set of all nodes, is the weight parameter in the first output layer, is the bias parameter in the first output layer, represents the sigmoid function;

[0138] In one embodiment of the present invention, the calculation formula of the second output layer is as follows:

[0139] ;

[0140] where represents the second output vector, and the i-th component of the second output vector represents the probability value of the i-th action; the set of all actions is called the action space, which contains all possible actions that can be selected;

[0141] In one embodiment of the present invention, an action is represented as a vector, and the j-th component of the vector represents the j-th representation parameter of the movement of the surface features of the glacier;

[0142] An action is represented as a matrix. The element in the first row and the L-th column of the matrix represents the representation parameter of the L-th displacement vector. The element in the second row and the L-th column represents the representation parameter of the L-th velocity field. The element in the third row and the L-th column represents the representation parameter of the L-th strain rate. The element in the fourth row and the L-th column represents the representation parameter of the L-th surface elevation change. The element in the fifth row and the L-th column represents the representation parameter of the L-th surface morphology change. represents the first hidden state of the v-th node in the second output layer of the input. represents the concatenation of the first hidden states of all nodes in the second output layer of the input. Z represents the set of nodes related to the recognition result. is the weight parameter in the second output layer. is the bias parameter in the second output layer. represents the sigmoid function;

[0143] In an embodiment of the present invention, the calculation formula of the second hidden layer is as follows:

[0144] ;

[0145] where represents the recognition feature of the second graph structure data of the d-th node in the l-th layer of the t-th sequence unit of the sequence data. represents the recognition feature of the second graph structure data of the u-th node in the t-th sequence unit of the sequence data in the l-1 layer. and respectively represent the sets of nodes with high association relationships with node d and node u. represents the cardinality of the set. represents the weight matrix of the second graph structure data in the l-th layer. , E represents the total number of layers. When l = 1 , represents the feature representation of the object indexed by the u-th node in the t-th sequence unit of the sequence data. is the sigmoid function.

[0146] In an embodiment of the present invention, the calculation formula of the fourth hidden layer is as follows:

[0147] ;

[0148] ;

[0149] where represents the fourth hidden state. represents the information of the movement mode of the input glacier. represents the first convolution function;

[0150] The calculation formula of the fifth hidden layer is as follows:

[0151] ;

[0152] ;

[0153] Among them, represents the fifth hidden state, represents the information of the movement of the surface features of the input glacier, represents the second convolution function.

[0154] In an embodiment of the present invention, the calculation formula of the first feature fusion layer is as follows:

[0155] ;

[0156] Among them, represents the first fusion state, represents the recognition feature of the second graph structure data of the last layer of the d-th node output when the t-th sequence unit of the sequence data is input to the second hidden layer, represents the set of all units of the second graph structure data, represents the splicing function, represents the summation weight matrix, represents the summation bias parameter.

[0157] In an embodiment of the present invention, the calculation formula of the fifth hidden layer is as follows:

[0158] ;

[0159] ;

[0160] ;

[0161] ;

[0162] Among them, , , represent the first, second, and third weight matrices in the fifth hidden layer, , , represent the first, second, and third bias parameters in the fifth hidden layer, represents the dot product, , and respectively represent the first, second, and third intermediate states of the fifth hidden layer, , and respectively represent the t-th and (t - 1)-th fifth hidden states, where n ≥ t ≥ 1, n represents the total number of sequence units of the sequence data, and when t = 1 , tanh is the hyperbolic tangent function, represents the sigmoid function.

[0163] In an embodiment of the present invention, the calculation formula of the third output layer is as follows:

[0164] ;

[0165] where represents the output matrix, and an element of the output matrix represents the change rate of the species core area of a species near a glacier, represents the g-th fifth hidden state, represents the concatenation of the fifth hidden states of all nodes, represents the set of all nodes, is the weight parameter in the third output layer, is the bias parameter in the third output layer, represents the S function.

[0166] In an embodiment of the present invention, the first loss function of the motion estimation model is as follows:

[0167] ;

[0168] where represents the first loss value, M represents the total number of motion pattern classification categories, represents the total number of training samples, is the sign value of the -th training sample. If the true motion pattern of the -th training sample is the c-th motion pattern classification category, its value is 1; otherwise, its value is 0. is the probability that the -th training sample predicted by the motion estimation model belongs to the c-th motion pattern classification category, represents the logarithmic function with the natural constant e as the base.

[0169] In an embodiment of the present invention, the second loss function of the habitat assessment model is as follows:

[0170] ;

[0171] where represents the second loss value, represents the total number of training samples, represents the true change rate of the species core area of the species near the glacier of the k-th sample, It represents the change rate of the species core area of the species near the glacier for the k-th sample output by the model.

[0172] As Figure 2 shown, in an embodiment of the present invention, an image data processing system based on the generation system technology of big data and artificial intelligence includes:

[0173] Data acquisition module 101: Collect motion estimation information and glacier animal habitat information based on the generation system technology of big data and artificial intelligence;

[0174] Data encoding module 102: Construct the first graph structure data, sequence data, and second graph structure data, including sequences or nodes and edges, as the input for establishing corresponding motion estimation models and habitat assessment models for each glacier, bedrock, habitat, and animal;

[0175] Data processing module 103: Input the encoded first graph structure data, sequence data, and second graph structure data into the motion estimation model and habitat assessment model, and output the results of the corresponding motion pattern classification and the results of the surface feature movement of the glacier;

[0176] Result evaluation module 104: Evaluate the dynamic characteristics, mass balance state, response to environmental changes, and impact on regional water resources of the glacier through the results of the corresponding motion pattern classification and the results of the surface feature movement of the glacier. At the same time, it can also combine the glacier animal habitat information to obtain the change rate of the species core area, and evaluate the impact of glacier changes on the species near the glacier.

[0177] At least one embodiment of the present disclosure provides a storage medium storing non-transitory computer-readable instructions for performing one or more steps in the foregoing image data processing method based on the generation system technology of big data and artificial intelligence.

[0178] Computer programs can be stored / distributed on suitable media, such as optical storage media or solid-state media supplied together with other hardware or as part of other hardware, but can also be distributed in other forms, such as via the Internet or other wired or wireless telecommunication systems. Any reference signs in the claims shall not be construed as limiting the scope.

[0179] The above describes the embodiments of this embodiment, but this embodiment is not limited to the above specific implementation manners. The above specific implementation manners are merely illustrative and not restrictive. Under the inspiration of this embodiment, those of ordinary skill in the art can also make many forms, all of which fall within the protection scope of this embodiment.

Claims

1. An image data processing method based on big data and artificial intelligence generation system technology, characterized in that, It includes the following steps: Step 100, collect motion estimation information and glacier animal habitat information; the motion estimation information includes optical and SAR satellite image information, satellite image imaging information, glacier information and bedrock information; the glacier animal habitat information includes animal species, animal habitat locations, animal habitat areas and animal numbers; Among them, the satellite image imaging information includes: satellite orbit parameters, imaging geometric information, and ground control point information of the imaging satellite; Glacier information: glacier boundary, glacier surface elevation, glacier surface velocity, glacier thickness, glacier mass balance, glacier surface temperature, glacier surface albedo and glacier surface morphology; Bedrock information: bedrock topography, bedrock elevation, bedrock slope, bedrock roughness, bedrock type, bedrock temperature and bedrock stress; Step 200, encode the motion estimation information into the first graph structure data. The first graph structure data includes nodes and edges connecting the nodes. A node represents a satellite image, a glacier, a bedrock or a satellite; There is an edge between the node representing the glacier in the same satellite image and the node representing the bedrock; There is an edge between the node representing the glacier in the satellite image and the node representing the glacier in the satellite image of the previous timestamp; If the similarity between the node representing the glacier in the satellite image and the node representing the glacier in the satellite image of the previous timestamp exceeds the set threshold, it is determined that there is an edge between them; If the similarity between the nodes representing the bedrock in the satellite image exceeds the set threshold, it is determined that there is an edge between them; There is an edge between the satellite image and the satellite; Construct the second graph structure data. The second graph structure data includes nodes and edges connecting the nodes. A node represents an animal or a habitat; There is an edge between the node representing an animal and all the nodes representing habitats. There is an edge between the nodes corresponding to all the animals in the same node representing a habitat. There is an edge between all the nodes representing habitats; Encode the glacier animal habitat information of each node into sequence data respectively. The sequence data includes n sequence units, and the t-th sequence unit represents the glacier animal habitat information of a node at the t-th moment; Step 300, input the first graph structure data into the motion estimation model. The motion estimation model includes a first hidden layer, a first output layer and a second output layer. The first hidden layer inputs the first graph structure data and outputs the first hidden state to the first output layer. The first output layer outputs the classification result of the glacier's motion pattern. The first hidden state is input to the second output layer, and the second output layer outputs the result of the movement of the glacier's surface features; The sequence data, the output result of the first output layer, and the output result of the second output layer are input into the habitat assessment model together. The habitat assessment model includes a second hidden layer, a third hidden layer, a fourth hidden layer, a fifth hidden layer, a first feature fusion layer, and a third output layer. Among them, the second hidden layer inputs the sequence data and outputs the second hidden state. The output results of the first output layer and the second output layer are respectively input into the fourth hidden layer and the fifth hidden layer. Then, the fourth hidden state and the fifth hidden state output by the fourth hidden layer and the fifth hidden layer are input into the first feature fusion layer together with the second hidden state. The first feature fusion layer outputs the first fusion state, and the output first fusion state is input into the third hidden layer. The third hidden layer outputs the third hidden state to the third output layer, and the third output layer outputs the result representing the change rate of the core area; The habitat assessment model runs twice. When running for the first time, the first feature fusion layer inputs the second hidden state and the default feature, and the size of the default feature is the same as that of the second hidden state. The first output layer outputs the initial change rate of the species core area; When running for the second time, the first feature fusion layer inputs the second hidden state, the fourth hidden state, and the fifth hidden state, and the third output layer outputs the change rate of the species core area for the current time; Step 400: According to the output result of the motion pattern classification and the result of the movement of the glacier's surface features, evaluate the sensitivity of the glacier to environmental changes, the stability of the glacier, and the potential risks, or the impact of glacier changes on regional water resources.

2. The image data processing method based on the generation system technology of big data and artificial intelligence according to claim 1, characterized in that, The calculation formula of the first hidden layer is as follows: ； Among them represents the first hidden state of the v-th node represents the graph weight parameter in the first hidden layer and respectively represent the node features of the v-th and u-th nodes of the first graph structure data is the set of nodes that have edges with node v of the first graph structure data represents the total number of nodes in ； wherein represents the aggregation coefficient between the v-th and u-th nodes, tanh represents the hyperbolic tangent function, and T represents the transpose, represents an adjustable parameter; The default value of is 1 / N, where N is the dimension of the node features; The calculation formula of the first output layer is as follows: ； wherein represents the first output vector, and the i-th component of the first output vector represents the probability that the movement pattern of the glacier belongs to the i-th state. represents the v-th first hidden state input to the first output layer. represents the concatenation of the first hidden states of all nodes in the input first output layer. represents the set of all nodes. is the weight parameter in the first output layer. is the bias parameter in the first output layer. represents the sigmoid function. The calculation formula of the second output layer is as follows: ； wherein represents a second output vector, and the i-th component of the second output vector represents the probability value of the i-th action; the set of all actions is called the action space, which contains all possible actions that can be selected; An action is represented as a vector, and the j-th component of the vector represents the j-th representation parameter of the movement of the glacier's surface features; An action is represented as a matrix. The element in the first row and the L-th column of the matrix represents the representation parameter of the L-th displacement vector. The element in the second row and the L-th column represents the representation parameter of the L-th velocity field. The element in the third row and the L-th column represents the representation parameter of the L-th strain rate. The element in the fourth row and the L-th column represents the representation parameter of the L-th surface elevation change. The element in the fifth row and the L-th column represents the representation parameter of the L-th surface morphology change. represents the first hidden state of the v-th node in the second output layer of the input. represents the concatenation of the first hidden states of all nodes in the second output layer of the input. Z represents the set of nodes related to the recognition result. is the weight parameter in the second output layer. is the bias parameter in the second output layer. represents the sigmoid function.

3. The image data processing method based on the technology of a generation system of big data and artificial intelligence according to claim 2, wherein The calculation formula of the second hidden layer is as follows: ； Among them, represents the second graph structure data recognition feature of the d-th node in the l-th layer of the t-th sequence unit of the sequence data, represents the second graph structure data recognition feature of the u-th node in the t-th sequence unit of the sequence data in the (l - 1)-th layer, and respectively represent the sets of nodes with high association relationships with nodes d and u, represents the cardinality of the set, represents the weight matrix of the second graph structure data in the l-th layer, , E represents the total number of layers, when l = 1 , represents the feature representation of the object with the u-th node index of the t-th sequence unit of the sequence data, is the sigmoid function; The calculation formula of the fourth hidden layer is as follows: ； ； Among them, represents the fourth hidden state, represents the information of the movement pattern of the input glacier, represents the first convolution function; The calculation formula of the fifth hidden layer is as follows: ； ； Among them, represents the fifth hidden state, represents the information of the movement of the surface features of the input glacier, represents the second convolution function.

4. An image data processing method based on the technology of a generation system of big data and artificial intelligence according to claim 3, characterized in that The calculation formula of the first feature fusion layer is as follows: ； Among them, represents the first fusion state, represents the recognition feature of the second graph structure data of the last layer of the d-th node output when the t-th sequence unit of the sequence data is input into the second hidden layer, represents the set of all units of the second graph structure data, represents the splicing function, represents the summation weight matrix, represents the summation bias parameter; The calculation formula of the fifth hidden layer is as follows: ； ； ； ； Among them, , , represent the first, second, and third weight matrices in the fifth hidden layer, , , represent the first, second, and third bias parameters in the fifth hidden layer, represents the dot product, , and respectively represent the first, second, and third intermediate states in the fifth hidden layer, , and respectively represent the t-th and (t - 1)-th fifth hidden states, where n ≥ t ≥ 1, n represents the total number of sequence units of the sequence data, and when t = 1 , tanh is the hyperbolic tangent function, represents the sigmoid function; The calculation formula of the third output layer is as follows: ； Among them represents the output matrix, and an element of the output matrix represents the change rate of the species core area of the species near a glacier represents the g-th fifth hidden state represents the concatenation of the fifth hidden states of all nodes represents the set of all nodes is the weight parameter in the third output layer is the bias parameter in the third output layer represents the sigmoid function 5. An image data processing system based on the technology of a generation system of big data and artificial intelligence, characterized in that, Execute the steps in an image data processing method of a generation system technology based on big data and artificial intelligence as described in any one of claims 1-4, including: Data acquisition module: Collect motion estimation information and glacier animal habitat information based on the generation system technology of big data and artificial intelligence; Data encoding module: Construct the first graph structure data, sequence data, and second graph structure data, including sequences or nodes and edges, and establish corresponding inputs for the motion estimation model and the habitat assessment model for each glacier, bedrock, habitat, and animal; Data processing module: Input the encoded first graph structure data, sequence data, and second graph structure data into the motion estimation model and the habitat assessment model, and output the corresponding results of the motion pattern classification and the results of the movement of the glacier's surface features; Result evaluation module: Through the corresponding results of the motion pattern classification and the results of the movement of the glacier's surface features, and also in combination with the glacier animal habitat information, obtain the change rate of the species core area.

6. A storage medium, characterized in that, Stored with non-transitory computer-readable instructions for performing the steps in the image data processing method of the big data and artificial intelligence-based generation system technology as described in any one of claims 1-4.

Images