An aeronautical magnetic anomaly signal data set construction and intelligent identification system

Abstract

The application discloses an aviation magnetic anomaly signal data set construction and intelligent identification system, relates to the field of aviation magnetic detection technology, and comprises a data selection and data set construction module, a training module and a test / identification module, wherein the data selection and data set construction module is used for arranging and target selecting magnetic data collected by a magnetic exploration airplane, so that samples with a standard size are generated and filled into a data set for model training or test; the training module supports multiple intelligent algorithms, is used for model training of the aviation magnetic anomaly signal data set, and monitors training and verification loss curves in the training process; and the test / identification module is used for executing anomaly signal identification by using a trained model. The application can continuously improve the performance of an identification algorithm according to actual collected data and adapt to magnetic detection requirements in different scenes.

Description

Technical Field

[0001] This invention relates to the field of airborne magnetic detection technology, and more specifically, to an airborne magnetic anomaly signal dataset construction and intelligent identification system. Background Technology

[0002] Airborne magnetic detection technology refers to the detection of ferromagnetic objects underground or underwater by mounting a magnetic detector on an aircraft platform. It is widely used in geological exploration, underwater target detection, and other scenarios. Airborne magnetic anomaly signal processing is one of the key technologies in airborne magnetic detection. It generally detects magnetic anomaly signals through standard orthogonal basis matched filtering or uses artificial intelligence algorithms to perform binary classification (recognition) of magnetic anomaly signals and interference / background signals. The effectiveness of intelligent recognition algorithms depends heavily on the quality of the dataset. A widely accepted conclusion is that "a good dataset + a standard artificial intelligence recognition algorithm" yields better recognition results than "a poor dataset + a high-performance artificial intelligence recognition algorithm." This fully demonstrates that a good dataset is crucial for improving the performance of recognition algorithms. A good dataset should be superior in both quantity and quality, meaning it should have a sufficiently large sample size and include samples under various conditions. To facilitate the construction of a dedicated dataset and the training and testing of intelligent models using this dataset, it is necessary to develop an airborne magnetic anomaly signal dataset construction and intelligent recognition system. Summary of the Invention

[0003] To address the need for aeromagnetic anomaly signal recognition technology in artificial intelligence models, this invention proposes an aeromagnetic anomaly signal dataset construction and intelligent recognition system. The system aims to select aeromagnetic anomaly signals and construct datasets, and to train and recognize artificial intelligence models.

[0004] To achieve the above technical objectives, this application provides a system for constructing and intelligently identifying airborne magnetic anomaly signal datasets, including: The data selection and dataset construction module is used to organize and select targets from the magnetic data collected by the magnetic probe aircraft to generate standard-sized samples, which are then added to the dataset for model training or testing. The training module supports a variety of intelligent algorithms for model training of aeromagnetic anomaly signal datasets and monitors the training and validation loss curves during the training process. The test / identification module is used to perform anomaly signal identification using a trained model.

[0005] Preferably, the data selection and dataset construction module is used to select waveform data for display within a total of 3 minutes before and after the alarm time; at the same time, it adjusts the filter frequency band according to the noise level of the data segment, obtains the wave width of the magnetic anomaly signal through the waveform diagram, and adjusts the flight altitude range and attitude to obtain negative sample data.

[0006] Preferably, the data selection and dataset construction module ensures that the training and test sets are not duplicated when generating the training and validation sets.

[0007] Preferably, the training module is used to set parameters such as batch size, learning rate, and number of training rounds during training, select a suitable artificial intelligence model for model training, and display the total number of samples, the number of positive and negative samples, and the number of target categories in the dataset. It also automatically saves the model parameters after training is completed for subsequent use.

[0008] Preferably, the test / identification module is used to perform abnormal signal identification using offline sample identification or real-time sliding window identification.

[0009] Preferably, the test / identification module is used to select a sample in the dataset for identification based on the trained model, and can display the waveform in real time and output the alarm location during identification.

[0010] Preferably, the test / identification module is also used to perform sliding window identification based on a long-term magnetic anomaly signal input from a trained model.

[0011] Preferably, the test / recognition module is used to select a single sample in the dataset, automatically recognize and output the result when performing offline sample recognition; when performing real-time sliding window recognition, it inputs a long-term signal acquired in real time, recognizes it segment by segment using a sliding window method; and outputs the target position and confidence level after the recognition is completed.

[0012] The present invention discloses the following technical effects: Standardized data processing and dataset construction: A dedicated sample generation process is designed for aeromagnetic anomaly signals, which can quickly generate standardized training / test samples, making it easy to build dedicated datasets according to specific needs; Integrated training-testing system: Enables a seamless end-to-end operation from dataset construction to model training to real-time recognition, eliminating the need to switch between multiple tools; Iterative optimization: It supports continuous improvement of the dataset and iterative training of the model, and can continuously improve the performance of the recognition algorithm based on actual collected data to adapt to the magnetic detection needs of different scenarios. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments 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.

[0014] Figure 1 This is a schematic diagram of the system structure described in this invention; Figure 2 This is a functional flowchart of the data selection and dataset construction module described in this invention; Figure 3 This is a flowchart of the training module workflow described in this invention; Figure 4 This is a flowchart of the testing / identification module described in this invention. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0016] like Figures 1-4 As shown, this invention provides a system for constructing and intelligently identifying aeromagnetic anomaly signal datasets. The system comprises three main functional modules, such as... Figure 1 As shown. The middle section is the data selection and dataset construction module, the left side is the training module, and the right side is the testing / recognition module, which specifically includes the following: 1. Data Selection and Dataset Construction Module: This module organizes and selects targets from the magnetic data collected by the magnetic anomaly aircraft to generate standard-sized samples, which are then added to the dataset for model training or testing. This module can display waveform data for a total of 3 minutes (adjustable) before and after the alarm time. Simultaneously, the filter band can be adjusted according to the noise level of this data segment, the waveform width of the magnetic anomaly signal can be obtained from the waveform graph, and negative sample data can be obtained by adjusting flight altitude range and attitude constraints. The sampling window size (tentatively 600 points), step size (tentatively 30 points), and initial point position can be set for display and sample saving. The storage locations of training and test samples can be modified as needed (note that training and test sets should not have duplicates).

[0017] 2. Training Module: The "Training Module" is used for model training of the aeromagnetic anomaly signal dataset and supports various intelligent algorithms. Users can select training parameters such as batch size and learning rate, choose the aeromagnetic anomaly dataset with specific component numbers, and select a specific artificial intelligence model (e.g., machine learning models such as neural networks, support vector machines, and random forests, and deep learning models such as CNNs and LSTMs). After selecting the parameters, clicking the "Start Training" button will begin training. The training process monitoring displays the training and validation loss curves in real time, facilitating monitoring the model's optimization towards the optimal direction. The dataset status display bar shows the total number of samples in the dataset, the number of positive and negative samples, and the number of target categories.

[0018] 3. Test / Recognition Module: This module is used for target recognition using trained models, supporting offline sample recognition and real-time sliding window recognition. The model selection button allows users to choose a trained model from the model library for target recognition. This module can select a specific sample from the dataset for recognition, or input a long-term magnetic anomaly signal for sliding window recognition. During recognition, the waveform can be displayed in real time, and the alarm location can be output.

[0019] Example: This invention discloses a system for constructing and intelligently identifying aeromagnetic anomaly signal datasets, comprising: System overall architecture and interface layout: The interface of this system is as follows Figure 1 As shown, the system comprises three parts: left, center, and right. The center part is the data selection and dataset construction module, used to select positive and negative samples from aeromagnetic anomaly signals and build a dedicated dataset. The left part is the training module; after selecting the AI ​​model and dataset, training begins by clicking the training button, and the model is saved after training. The right part is the testing / recognition module, which, after loading the trained model parameters, identifies the input magnetic anomaly signals. This system can complete the entire process from dataset construction to model training and recognition in one integrated manner, without the need to switch software.

[0020] The implementation process of the data selection and dataset construction module is as follows: Figure 2 As shown, the specific steps are as follows: 1. Raw data input: Input the airborne magnetic anomaly signal into the system; 2. Target time selection: The operator selects the target time, and the system automatically retrieves and displays waveform data from 3 minutes before and after that time (adjustable); 3. Filtering: The operator can manually adjust the filter band to suppress noise; 4. Positive and negative sample generation: The operator selects positive samples based on the location of the magnetic anomaly signal and selects background data as negative samples; 5. Sample parameter settings: Set the sampling window size (default 600 data points, adjustable), step size (default 30 data points, adjustable), and initial point position, and the system automatically generates positive and negative samples; 6. Dataset storage: The operator sets the storage paths for training and testing samples to complete the dataset construction.

[0021] The implementation process of the training module is as follows: Figure 3 As shown, the specific steps are as follows: 1. Dataset Import: Import the generated dataset into the training module; 2. Parameter Settings: Set parameters such as batch size, learning rate, and number of training epochs during training; 3. Model Selection: Select a suitable artificial intelligence model, such as machine learning models like random forests or deep learning models like CNNs; 4. Training Start and Monitoring: Click the "Start Training" button to start training, which will display the training loss curve; 5. Dataset Status Display: Display information such as the total number of samples, the number of positive and negative samples, and the number of target classes in the dataset; 6. Model Storage: Automatically save the model parameters after training is complete for later retrieval.

[0022] The implementation process of the testing / identification module is as follows: Figure 4 As shown, the specific process is as follows: 1. Model call: Load the parameters of the trained artificial intelligence model; 2. Recognition mode selection: When performing offline sample recognition, first select a single sample in the dataset, and the intelligent model will automatically recognize and output the result; when performing real-time sliding window recognition, input a long-term signal collected in real time, and the system will recognize it segment by segment using the sliding window method; 3. Result output: After the recognition is completed, output the target location and confidence level.

[0023] In summary, this invention designs a dedicated sample generation process for aero-magnetic anomaly signals, which can quickly generate standardized training / test samples, facilitating the construction of dedicated datasets according to specific needs. Furthermore, this invention achieves a fully integrated operation of dataset construction, model training, and real-time recognition, eliminating the need to switch between multiple tools. Moreover, this invention supports continuous improvement of the dataset and iterative training of the model, continuously enhancing the performance of the recognition algorithm based on actual collected data, and adapting to the magnetic detection needs of different scenarios.

[0024] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0025] In the description of this invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0026] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A system for constructing and intelligently identifying aeromagnetic anomaly signal datasets, characterized in that, include: The data selection and dataset construction module is used to organize and select targets from the magnetic data collected by the magnetic probe aircraft to generate standard-sized samples, which are then added to the dataset for model training or testing. The training module supports a variety of intelligent algorithms for model training of aeromagnetic anomaly signal datasets and monitors the training and validation loss curves during the training process. The test / identification module is used to perform anomaly signal identification using a trained model.

2. The system for constructing and intelligently identifying an aeromagnetic anomaly signal dataset according to claim 1, characterized in that: The data selection and dataset construction module is used to select waveform data for display within a total of 3 minutes before and after the alarm time; at the same time, it adjusts the filter frequency band according to the noise level of the data segment, obtains the wave width of the magnetic anomaly signal through the waveform diagram, and adjusts the flight altitude range and attitude to obtain negative sample data.

3. The system for constructing and intelligently identifying aeromagnetic anomaly signal datasets according to claim 2, characterized in that: The data selection and dataset construction module ensures that the training and test sets are not duplicated when generating the training and validation sets.

4. The system for constructing and intelligently identifying an aeromagnetic anomaly signal dataset according to claim 3, characterized in that: The training module is used to set parameters such as batch size, learning rate, and number of training rounds during training, select an appropriate artificial intelligence model for model training, and display the total number of samples, the number of positive and negative samples, and the number of target categories in the dataset. It also automatically saves the model parameters after training is completed for subsequent use.

5. The system for constructing and intelligently identifying an aeromagnetic anomaly signal dataset according to claim 4, characterized in that: The test / identification module is used to perform abnormal signal identification using offline sample identification or real-time sliding window identification.

6. The system for constructing and intelligently identifying an aeromagnetic anomaly signal dataset according to claim 5, characterized in that: The test / recognition module is used to select a sample in the dataset for recognition based on the trained model. During recognition, the waveform can be displayed in real time and the alarm location can be output.

7. The system for constructing and intelligently identifying an aeromagnetic anomaly signal dataset according to claim 6, characterized in that: The test / identification module is also used to perform sliding window identification based on a long-term magnetic anomaly signal inputted by the trained model.

8. The system for constructing and intelligently identifying an aeromagnetic anomaly signal dataset according to claim 7, characterized in that: The test / recognition module is used to select a single sample in the dataset for offline sample recognition, automatically recognize and output the result; when performing real-time sliding window recognition, it inputs a long-term signal acquired in real time, recognizes it segment by segment using a sliding window method; and outputs the target position and confidence level after the recognition is completed.

Images