Big data-based method and system for assessing reliability of basin glacier change information

Abstract

The present application relates to the technical field of electric digital data processing, and especially relates to a watershed glacier change information reliability evaluation method and system based on big data, which comprises the following steps: obtaining an annotated and sorted target region glacier standard form atlas; performing segmentation on the standard form atlas to obtain a training set, a first test set and a second test set; uploading the training set and the first test set to hierarchical iterative model training to obtain a type 1 evaluation model; uploading the training set and the second test set to hierarchical iterative model training to obtain a type 2 evaluation model; fusing the type 1 evaluation model and the type 2 evaluation model to obtain a reliability evaluation model; and performing reliability evaluation on the watershed glacier change information according to the reliability evaluation model. The present application realizes reliability evaluation on the watershed glacier change information.

Description

Technical Field

[0001] This invention relates to the field of electronic digital data processing technology, and in particular to a method for assessing the reliability of watershed glacier change information based on big data, and also to an assessment system. Background Technology

[0002] Glaciers are a crucial component of the cryosphere. They are not only a significant driver of climate change but also serve as recorders and early warning systems for it. Therefore, rapidly and accurately detecting areas of glacier change is of paramount importance. Current methods for glacier change detection based on remote sensing imagery mainly include simple algebraic algorithms, image classification, and structural feature analysis. However, due to continuous global warming and accelerated glacier melting, glacier locations are constantly shifting. Existing technologies are no longer adequate for effectively predicting the changing locations of glaciers, leading to gaps and uncontrollable risks in global climate change forecasting.

[0003] The existing Chinese invention patent application, published on March 21, 2023, with publication number CN115830466A, entitled "A Remote Sensing Detection Method for Glacier Change Based on Deep Siamese Neural Network," discloses the following steps: preprocessing dual-temporal glacier remote sensing images to obtain training data for glacier change detection; inputting the dual-temporal training data into a deep Siamese neural network for training to obtain a trained deep Siamese neural network; inputting dual-temporal glacier remote sensing image data into the trained deep Siamese neural network to obtain a binary prediction image for glacier change detection; and vectorizing the obtained binary prediction image for glacier change detection to obtain a vector map of the glacier change region.

[0004] The aforementioned technologies can effectively enhance the spatial characteristics of areas with subtle changes, obtain glacier remote sensing image change detection maps, and improve the accuracy of glacier change detection. However, they cannot effectively integrate and analyze multiple data sets, making it difficult to comprehensively and accurately assess the reliability of glacier change information. Summary of the Invention

[0005] The inventors discovered through research that glaciers, as sensitive indicators of global climate change, have a profound impact on sea-level rise, water resource supply, and ecosystem balance. Accurately acquiring and assessing information on glacier changes in watersheds is crucial for predicting climate change trends, rationally planning water resource utilization, and protecting the ecological environment. Traditional glacier monitoring methods, such as on-site measurements, can obtain relatively accurate data, but are limited by geographical conditions, resulting in a limited measurement range and low efficiency. While remote sensing monitoring methods can achieve large-area monitoring, they have limitations in terms of accuracy and capturing detailed changes in glaciers.

[0006] The purpose of this invention is to provide a method and system for reliability assessment of watershed glacier change information based on big data. By fusing a type I assessment model and a type II assessment model to obtain a reliability assessment model, the reliability of watershed glacier change information can be assessed, thereby solving the technical problem that existing technologies cannot provide a reliable assessment of multiple data in glacier change.

[0007] According to one aspect of the present invention, a method for reliability assessment of watershed glacier change information based on big data is provided, the method being executed by a processor.

[0008] Obtain a standard morphological atlas of glaciers in the target area that has been labeled and sorted;

[0009] Segment the standard morphological image set to obtain the training set, the first test set, and the second test set;

[0010] Upload the training set and the first test set to the hierarchical iterative model training to obtain a type I evaluation model;

[0011] Upload the training set and the second test set to the hierarchical iterative model for training to obtain the type II evaluation model;

[0012] By integrating the Type I assessment model and the Type II assessment model, a reliability assessment model is obtained.

[0013] Based on the reliability assessment model, a reliability assessment of the watershed glacier change information was performed.

[0014] In some embodiments, the process of obtaining a standard morphological atlas of glaciers for a labeled and sorted target region is as follows:

[0015] Connect to high-resolution satellites to acquire at least images of the standard area, standard outline, and conventional surface features of glaciers in the target area;

[0016] Bubble sorting was performed on images of glacier standard area, standard outline, and conventional surface features to obtain a time-series image sequence from near to far.

[0017] Grouping is performed on glacier standard area, standard contour, and conventional surface feature images from the same time series;

[0018] Data cleaning was performed on each set of images to obtain a standard morphological atlas of glaciers in the target area.

[0019] In some embodiments, the process of segmenting the standard morphological image set to obtain the training set, the first test set, and the second test set is as follows:

[0020] Based on the machine learning algorithm, the standard morphological map of the target area glacier is divided into a fixed ratio to obtain a training set, a first test set, and a second test set, wherein the ratio of the training set, the first test set, and the second test set is at least 5:4:1.

[0021] In some embodiments, the process of uploading the training set and the first test set to the hierarchical iterative model for training to obtain an evaluation model is as follows:

[0022] Upload the training set to the hierarchical iterative model and perform optimization;

[0023] The first test set is uploaded to the optimized hierarchical iterative model according to the sample definition. The parameters of the hierarchical iterative model are adjusted to obtain the optimal hierarchical iterative model.

[0024] Uncertainty analysis is performed on the optimal hierarchical iterative model to obtain an evaluation model.

[0025] In some embodiments, the process of uploading the training set and the second test set to the hierarchical iterative model for training to obtain the type II evaluation model is as follows:

[0026] Upload the training set to the hierarchical iterative model and perform optimization;

[0027] The second test set is uploaded to the optimized hierarchical iterative model according to the sample definition. The parameters of the hierarchical iterative model are adjusted to obtain the optimal hierarchical iterative model.

[0028] Uncertainty analysis is performed on the optimal hierarchical iterative model to obtain a type II evaluation model.

[0029] In some embodiments, the process of fusing a type-one evaluation model with a type-two evaluation model to obtain a reliability evaluation model is as follows:

[0030] The execution data from the first test set and the second test set are merged to obtain the merged test set;

[0031] Data cleaning is performed on the merged test set to obtain an interference-free test set. This test set is then used as the training set and uploaded to both the Type I and Type II evaluation models to obtain the first evaluation data.

[0032] Construct a meta-model, use the first evaluation data as the input layer features, perform training, and obtain the second evaluation data;

[0033] Compare the first evaluation data with the second evaluation data to obtain the optimal evaluation data;

[0034] The optimal evaluation data is used as the training set, uploaded to the meta-model, and the parameters are optimized to obtain the reliability evaluation model.

[0035] In some embodiments, the process of performing a reliability assessment of watershed glacier change information according to a reliability assessment model is as follows:

[0036] Acquire image data of glacier characterization changes in the target area at different time series, and perform a reliability assessment of the watershed glacier change information based on a reliability assessment model.

[0037] According to another aspect of the present invention, a reliability assessment system for watershed glacier change information based on big data is provided. The system includes a processor and further comprises:

[0038] The acquisition module is used to obtain a standard morphological atlas of glaciers in the target area that has been labeled and sorted.

[0039] The preprocessing module is used to perform segmentation on the standard morphological image set to obtain the training set, the first test set, and the second test set.

[0040] The first training module is used to upload the training set and the first test set to the hierarchical iterative model training to obtain an evaluation model.

[0041] The second training module is used to upload the training set and the second test set to the hierarchical iterative model training to obtain the type II evaluation model.

[0042] The merged training module is used to fuse the Type I evaluation model and the Type II evaluation model to obtain a reliability evaluation model.

[0043] The execution module is used to perform a reliability assessment of watershed glacier change information based on the reliability assessment model.

[0044] In some embodiments, the processor data connection acquisition module, the preprocessing module, the first training module, the second training module, the merged training module, and the execution module are included.

[0045] Compared with existing technologies, this invention has the following advantages and beneficial effects: obtaining a labeled and sorted standard morphological map set of glaciers in the target area can improve the accuracy of reliability assessment in the method of this invention; segmenting the standard morphological map set to obtain a training set, a first test set, and a second test set, unlike traditional machine learning segmentation methods, adopting a more detailed test set division to achieve precision in model training; uploading the training set and the first test set to a hierarchical iterative model for training to obtain a type I evaluation model; uploading the training set and the second test set to a hierarchical iterative model for training to obtain a type II evaluation model, obtaining different iterative training models by training on two different datasets, achieving matching of multiple data information in glacier changes; fusing the type I evaluation model and the type II evaluation model to obtain a reliability assessment model, performing reliability assessment of watershed glacier change information based on the reliability assessment model, and fusing the type I and type II evaluation models to obtain a reliability assessment model that achieves a more comprehensive and accurate analysis of glacier change information, effectively solving the problems of incomplete and inaccurate results caused by single data processing in existing technologies. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 This is a flowchart of the evaluation method of the present invention;

[0048] Figure 2 This is a structural diagram of the evaluation system of the present invention;

[0049] Figure 3 This is a diagram illustrating the application scenario of the evaluation system of this invention. Detailed Implementation

[0050] The following will refer to the appendices in the embodiments of the present invention. Figure 1-3 The technical solutions in the embodiments of the present invention will be clearly and completely described together. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Example

[0051] Figure 1The flowchart illustrates the reliability assessment method for watershed glacier change information based on big data provided in this embodiment. This method is executed by a processor, which can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0052] The methods specifically include:

[0053] Obtain a labeled and sorted standard morphological atlas of glaciers in the target area. Using the characteristic images of glaciers in the target area as image sources, arrange the characteristic images temporally based on bubble sort to obtain time-series labeled images in ascending order of time, which facilitates image management. It can be understood that time-series labeled images can help models or observers find standard characteristic images of glaciers at different time points more quickly and accurately.

[0054] In some possible implementations, high-resolution satellites are connected to acquire images of at least the standard area, standard outline, and conventional surface features of glaciers in the target area. For example, high-resolution images from Landsat satellites at different times can be acquired, providing at least images of the standard area, standard outline, and conventional surface features of glaciers in the target area. These images allow for a direct observation of changes in glacier boundaries. Simultaneously, synthetic aperture radar (SAR) satellite data is used to obtain information on changes in the internal structure and thickness of the glacier. Next, bubble sorting is performed on the images of the standard area, standard outline, and conventional surface features of the glacier to obtain a time-series image sequence from recent to distant. Images of the same time sequence are grouped. In this embodiment, images of different glaciers' standard area, standard outline, and conventional surface features are classified based on the same time period, with at least images from the previous 5 years being classified. It should be noted that for glaciers in geologically active areas, researchers may conduct short-term image acquisitions of 3-5 years to quickly assess the impact of geological changes on the glacier; therefore, this embodiment selects a period of at least 5 years. Finally, data cleaning is performed on each group of images to obtain a standard morphological atlas of glaciers in the target area. In this embodiment, data cleaning includes at least removing noisy data, outliers, and duplicate data. For example, for abnormal pixels in satellite remote sensing data caused by cloud cover or sensor malfunction, interpolation algorithms are used for repair; for abrupt values ​​in ground monitoring data, they are corrected or removed by comparing and analyzing data from adjacent time periods.

[0055] The standard morphological image set is segmented to obtain a training set, a first test set, and a second test set. Here, the segmentation principle is based on the training and test set segmentation principle in machine learning, but this embodiment optimizes the segmentation ratio to further improve the accuracy of the final reliability assessment.

[0056] In some possible implementations, based on machine learning algorithms, the standard morphological atlas of glaciers in the target region is divided into a fixed-ratio data set to obtain a training set, a first test set, and a second test set, wherein the ratio of the training set, the first test set, and the second test set is at least 5:4:1. The 4:1 ratio of the first test set to the second test set is because, in practical work, the inventors found that the testing of general models, i.e., parameter optimization, is very limited to a single data set and cannot be optimized multiple times. This leads to high uncertainty in the model optimization results. Using a 4:1 ratio can largely avoid this problem. Of course, when dealing with other possible areas and situations, different ratios can be used to adapt to other training environments.

[0057] The training set and the first test set are uploaded to the hierarchical iterative model for training to obtain a Type I evaluation model. The training set and the second test set are then uploaded to the hierarchical iterative model for training to obtain a Type II evaluation model. By uploading the first and second test sets respectively, different evaluation models are obtained. It can be understood that the Type I and Type II evaluation models have different evaluation focuses. Due to the different test sets, the training focus of the models also differs, thus they are complementary. However, this differs from existing models trained on a single test set. Existing unified training sets do not focus on any particular point but rather present a uniform effect. This often results in the application of the trained model not having a significant corresponding advantage. In the glacier change scenario of this embodiment, the model has a specific focus, which greatly improves the accuracy of subsequent reliability assessment results.

[0058] In some possible implementations, the training set is uploaded to a hierarchical iterative model for optimization.

[0059] The first test set is uploaded to the optimized hierarchical iterative model according to the sample definition. The parameters of the hierarchical iterative model are adjusted to obtain the optimal hierarchical iterative model. Uncertainty analysis is performed on the optimal hierarchical iterative model to obtain a type I evaluation model. Similarly, the training set is uploaded to the hierarchical iterative model and optimized. The second test set is uploaded to the optimized hierarchical iterative model according to the sample definition. The parameters of the hierarchical iterative model are adjusted to obtain the optimal hierarchical iterative model.

[0060] Uncertainty analysis was performed on the optimal hierarchical iterative model to obtain a type II evaluation model. Considering the inherent uncertainties in glacier change data collection, this embodiment performs uncertainty analysis on the optimal hierarchical iterative model, analyzing the fluctuation range and confidence interval of the results through multiple simulations. For example, this embodiment uses the Monte Carlo simulation method to randomly perturb the standard glacier morphology atlas of the target area, runs the optimal hierarchical iterative model multiple times, and statistically analyzes the distribution of the results to obtain the uncertainty range of the results.

[0061] A reliability assessment model is obtained by fusing a type-one assessment model with a type-two assessment model. Model fusion combines the advantages of multiple models to improve the model's performance and generalization ability. Common implementation methods include averaging, voting, stacking, and fusion trees. This embodiment uses the stacking method to fuse the type-one and type-two assessment models to obtain the reliability assessment model.

[0062] In some possible implementations, the first test set and the second test set are merged to obtain a merged test set; the merged test set is cleaned to obtain an interference-free test set; this test set is then used as a training set and uploaded to both the Type I and Type II evaluation models to obtain the first evaluation data.

[0063] A meta-model is constructed, and the first evaluation data is used as the input layer features to be trained to obtain the second evaluation data. The first evaluation data and the second evaluation data are compared to obtain the optimal evaluation data. The optimal evaluation data is used as the training set, uploaded to the meta-model, and the parameters are optimized to obtain the reliability evaluation model.

[0064] A metamodel is a model of models, defining the rules, concepts, and relationships followed when constructing a specific model. Since a complex system or domain may contain multiple different models, a metamodel provides a unified standard and specification, ensuring consistency and compatibility between models. Common metamodels include the ER metamodel, UML metamodel, and MOF metamodel. This embodiment uses the MOF metamodel, taking the first evaluation data as input to the MOF metamodel. The metamodel determines the standard, i.e., the evaluation criterion, and then uses this criterion as the output, comparing it with the first evaluation data. The more complete and accurate data is selected as the optimal evaluation data. For example, if the first evaluation data includes a glacier corner collapse area of ​​2 square meters and a thickness variation of 50 centimeters, and the second evaluation data includes a glacier corner collapse area of ​​approximately 2 square meters and a thickness variation of approximately 50 centimeters, the first evaluation data is determined to be the optimal evaluation data because it is more accurate, and this process is repeated for each evaluation. The optimal evaluation data is used as the training set and uploaded to the meta-model. The optimization logic of the machine learning model is referenced, and the corresponding parameter optimization is performed to obtain the reliability evaluation model.

[0065] Based on the reliability assessment model, a reliability assessment of the watershed glacier change information was performed.

[0066] In some possible implementations, image data depicting glacier changes at different time points in the target area are acquired, and a reliability assessment of the watershed glacier change information is performed based on a reliability assessment model. It should be noted that this reliability assessment model performs a reliability assessment of the corresponding watershed glacier change information for the target area. Example

[0067] Figure 2This is a structural diagram of the evaluation system of the present invention. Exemplarily, the method can be divided into one or more modules, one or more of which are stored in memory and executed by a processor to complete the present invention. One or more modules can be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program. For example, the computer program can be divided into an acquisition module, a preprocessing module, a first training module, a second training module, a merging training module, and an execution module. The specific functions of each module are as follows: the acquisition module is used to obtain a labeled and sorted standard morphological map set of glaciers in the target area; the preprocessing module is used to segment the standard morphological map set to obtain a training set, a first test set, and a second test set; the first training module is used to upload the training set and the first test set to a hierarchical iterative model for training to obtain a type I evaluation model; the second training module is used to upload the training set and the second test set to a hierarchical iterative model for training to obtain a type II evaluation model; the merging training module is used to fuse the type I evaluation model and the type II evaluation model to obtain a reliability evaluation model; the execution module is used to perform a reliability evaluation of the watershed glacier change information according to the reliability evaluation model.

[0068] In some possible implementations, the processor data connection acquisition module, the preprocessing module, the first training module, the second training module, the merged training module, and the execution module are included.

[0069] The specific example of the big data-based watershed glacier change information reliability assessment method in the aforementioned embodiment 1 is also applicable to the big data-based watershed glacier change information reliability assessment system in this embodiment. Through the foregoing detailed description of the big data-based watershed glacier change information reliability assessment method, those skilled in the art can clearly understand the big data-based watershed glacier change information reliability assessment system in this embodiment. Therefore, for the sake of brevity, it will not be described in detail here.

[0070] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered illustrative and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the scope of the invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0071] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A method for evaluating the reliability of basin glacier change information based on big data, the method being executed by a processor and characterized by, obtaining a labeled and sorted standard glacier morphology atlas of a target region; performing segmentation on the standard morphology atlas to obtain a training set, a first test set, and a second test set; uploading the training set and the first test set to hierarchical iterative model training to obtain a first evaluation model; uploading the training set and the second test set to hierarchical iterative model training to obtain a second evaluation model; fusing the first evaluation model and the second evaluation model to obtain a reliability evaluation model; performing reliability evaluation of basin glacier change information according to the reliability evaluation model; the process of fusing the first evaluation model and the second evaluation model to obtain the reliability evaluation model is: performing data fusion on the first test set and the second test set to obtain a merged test set; performing data cleaning on the merged test set to obtain an interference-free test set, and uploading the test set as a training set to the first evaluation model and the second evaluation model to obtain first evaluation data; constructing a meta-model, inputting the first evaluation data as a feature to be input to the input layer, performing training, and obtaining second evaluation data; comparing the first evaluation data and the second evaluation data to obtain optimal evaluation data; uploading the optimal evaluation data as a training set to the meta-model to optimize parameters and obtain the reliability evaluation model.

2. The method of claim 1, wherein, the process of obtaining a labeled and sorted standard glacier morphology atlas of a target region is: connecting high-resolution satellites to obtain at least standard glacier area, standard contour, and conventional surface feature images of the target region; performing bubble sort on the standard glacier area, standard contour, and conventional surface feature images to obtain time sequence images from near to far; grouping the same time sequence of standard glacier area, standard contour, and conventional surface feature images; performing data cleaning on each group of images to obtain a standard glacier morphology atlas of the target region.

3. The method of claim 2, wherein, the process of performing segmentation on the standard morphology atlas to obtain a training set, a first test set, and a second test set is: performing fixed-ratio data segmentation on the target region glacier standard morphology atlas according to a machine learning algorithm to obtain a training set, a first test set, and a second test set, wherein the ratio of the training set, the first test set, and the second test set is at least 5:4:

1.

4. The method of claim 3, wherein, the process of uploading the training set and the first test set to hierarchical iterative model training to obtain a first evaluation model is: uploading the training set to the hierarchical iterative model to perform optimization; uploading the first test set to the optimized hierarchical iterative model according to sample definition to adjust the hierarchical iterative model parameters to obtain an optimal hierarchical iterative model; performing uncertainty analysis on the optimal hierarchical iterative model to obtain a first evaluation model.

5. The method of claim 4, wherein, the process of uploading the training set and the second test set to hierarchical iterative model training to obtain a second evaluation model is: uploading the training set to the hierarchical iterative model to perform optimization; uploading the second test set to the optimized hierarchical iterative model according to sample definition to adjust the hierarchical iterative model parameters to obtain an optimal hierarchical iterative model; performing uncertainty analysis on the optimal hierarchical iterative model to obtain a second evaluation model.

6. The method of claim 1, wherein, The process of performing reliability evaluation of the basin glacier change information according to the reliability evaluation model is: Obtain glacier change image data corresponding to different time sequences of the target region, and perform reliability evaluation of the basin glacier change information based on the reliability evaluation model.

7. A system for reliability assessment of basin glacier change information based on big data, the system comprising a processor, characterized in that, Also includes: The acquisition module is used to obtain the annotated and sorted target region glacier standard morphology atlas set; The preprocessing module is used to perform segmentation on the standard morphology atlas set to obtain a training set, a first test set, and a second test set; The first training module is used to upload the training set and the first test set to the hierarchical iterative model training to obtain a type evaluation model; The second training module is used to upload the training set and the second test set to the hierarchical iterative model training to obtain a type evaluation model; The merging training module is used to fuse the type evaluation model and the type evaluation model to obtain a reliability evaluation model; wherein the process of fusing the type evaluation model and the type evaluation model to obtain the reliability evaluation model is: performing data fusion on the first test set and the second test set to obtain a merged test set; performing data cleaning on the merged test set to obtain an interference-free test set, uploading the test set as a training set to the type evaluation model and the type evaluation model to obtain first evaluation data; constructing a meta model, inputting the first evaluation data as input layer features, performing training to obtain second evaluation data; comparing the first evaluation data and the second evaluation data to obtain optimal evaluation data; uploading the optimal evaluation data as a training set to the meta model to optimize parameters to obtain the reliability evaluation model; The execution module is used to perform reliability evaluation of the basin glacier change information according to the reliability evaluation model.

8. The system of claim 7, wherein, The processor is data-connected with the acquisition module, the preprocessing module, the first training module, the second training module, the merging training module, and the execution module.

Images