A geomagnetic adaptation area positioning method and device for geomagnetic navigation

Abstract

The present specification discloses a method and device for positioning a geomagnetic adaptation area for geomagnetic navigation, specifically comprising: determining a geomagnetic feature map corresponding to the target area according to the geomagnetic information map corresponding to the target area. The geomagnetic feature map is input into a scale division model, and a geomagnetic feature map corresponding to the geomagnetic information map at each image resolution is generated by the scale division model. For each scale feature map, a candidate geomagnetic adaptation area in the target area at the corresponding image resolution is determined. According to each candidate geomagnetic adaptation area, the geomagnetic adaptation area of ​​the target area is determined. Compared with the current technology of manually determining the geomagnetic adaptation area, the method of the present specification is more intelligent and more automated. While the efficiency and accuracy of determining the geomagnetic adaptation area are improved, the consumption of human resources and time costs is also greatly saved, so that the execution efficiency of subsequent geomagnetic route planning based on the geomagnetic adaptation area is also improved.

Description

Technical Field

[0001] The present invention relates to the field of geomagnetic navigation, and in particular to a method and device for positioning a geomagnetic adaptation area for geomagnetic navigation. Background Art

[0002] With the rapid development of science and technology, today's navigation technology has become increasingly mature and widely used and popularized. This technology has brought great convenience to various application fields and people's daily lives. In the current navigation field, geomagnetic matching navigation is a special navigation technology that uses the characteristics of the earth's magnetic field. Compared with other types of navigation technologies, this technology has the advantage of unlimited application areas and can be used in a variety of environments such as air, land and underwater. It is not easily affected by electromagnetic interference during application, and there is no navigation error that gradually accumulates in the inertial navigation system. Reasonable selection of the geomagnetic adaptation area based on geomagnetic navigation can better improve the navigation accuracy and practicality of geomagnetic navigation technology.

[0003] Regarding the selection of geomagnetic adaptation areas, the current technology mainly uses visual manual interpretation of geomagnetic data collected in the target area or manual screening of statistical parameters of some geomagnetic data as the basis for distinguishing geomagnetic features, thereby screening out range areas with obvious geomagnetic changes and numerical differences in the manually divided candidate areas as geomagnetic adaptation areas. This method is highly subjective, has a high missed detection rate, and the efficiency of determining geomagnetic adaptation areas is relatively low. When determining the geomagnetic adaptation area corresponding to a large target area, a large amount of human resources and time costs are required, which seriously affects the efficiency of determining the geomagnetic adaptation area, as well as the speed of advancing subsequent target tasks such as track planning based on the geomagnetic adaptation area.

[0004] Therefore, how to efficiently and accurately determine the geomagnetic adaptation area in the target area that can be used for trajectory planning is an urgent problem to be solved. Summary of the invention

[0005] This specification provides a geomagnetic adaptation area positioning method and device for geomagnetic navigation to partially solve the above-mentioned problems existing in the prior art.

[0006] This manual adopts the following technical solutions:

[0007] This specification provides a geomagnetic adaptation area positioning method for geomagnetic navigation, including:

[0008] Obtaining a geomagnetic information map corresponding to the target area, the geomagnetic information map is used to characterize the geomagnetic distribution in the target area, and for each grid point preset in the target area, the geomagnetic information map records geomagnetic data corresponding to the grid point, and the geomagnetic data corresponding to the grid point at least includes three-dimensional geomagnetic vector data corresponding to the grid point or combined data composed of the total geomagnetic field strength, magnetic declination and magnetic inclination corresponding to the grid point;

[0009] According to the geomagnetic information map, determine the geomagnetic characteristic map corresponding to the target area, and input the geomagnetic characteristic map into a preset scale division model, so that the scale division model generates each scale characteristic map corresponding to the geomagnetic information map according to the geomagnetic characteristic map, and for each scale characteristic map, the scale characteristic map is used to represent the geomagnetic characteristic map of the geomagnetic information map at the image resolution corresponding to the scale characteristic map;

[0010] For each scale characteristic map, determining a candidate geomagnetic adaptation region in the target area at an image resolution corresponding to the scale characteristic map;

[0011] According to the selected geomagnetic adaptation areas corresponding to the scale feature maps, the geomagnetic adaptation area corresponding to the target area is determined.

[0012] Optionally, determining a geomagnetic characteristic map corresponding to the target area according to the geomagnetic information map specifically includes:

[0013] The geomagnetic information map is input into a pre-trained feature extraction model, so that the feature extraction model determines the geomagnetic feature map corresponding to the target area based on the geomagnetic information map.

[0014] Optionally, training the feature extraction model specifically includes:

[0015] Get sample geomagnetic information map;

[0016] Performing image enhancement processing on the sample geomagnetic information map to obtain a global enhanced image corresponding to the sample geomagnetic information map, and performing image enhancement processing on the sample local geomagnetic information map corresponding to the sample geomagnetic information map to obtain a local enhanced image corresponding to the sample geomagnetic information map, wherein the image enhancement processing is at least any one of image size cropping processing, rotation processing and flipping processing, the sample local geomagnetic information map is obtained by first performing the image enhancement processing on the sample geomagnetic information map, and then dividing the sample geomagnetic information map after the image enhancement processing according to a preset dividing ratio, and the local enhanced image is obtained by splicing the sample local geomagnetic information maps corresponding to the sample geomagnetic information map according to the original image position;

[0017] Input the global enhanced image and the local enhanced image into the feature extraction model to be trained, so that the feature extraction model to be trained determines the global feature map corresponding to the sample geomagnetic information map through a preset global encoder according to the global enhanced image, and determines the global momentum feature map corresponding to the sample geomagnetic information map through a preset global momentum encoder, and determines the local feature map corresponding to the sample geomagnetic information map through a preset local encoder, and determines the local momentum feature map corresponding to the sample geomagnetic information map through a preset local momentum encoder, wherein the model parameters of the global momentum encoder are updated with momentum according to the model parameters of the global encoder at a preset update ratio, and the model parameters of the local momentum encoder are updated with momentum according to the model parameters of the local encoder at a preset update ratio;

[0018] Determine a global loss value corresponding to the feature extraction model to be trained according to the global feature map and the global momentum feature map, and determine a local loss value corresponding to the feature extraction model to be trained according to the local feature map and the local momentum feature map, wherein the size of the global loss value is negatively correlated with the similarity between the global feature map and the global momentum feature map, and the size of the local loss value is negatively correlated with the similarity between the local feature map and the local momentum feature map;

[0019] According to the global loss value and the local loss value, a total loss value corresponding to the feature extraction model to be trained is determined, and according to the total loss value, model training is performed on the feature extraction model to be trained.

[0020] Optionally, before determining a total loss value corresponding to the feature extraction model to be trained according to the global loss value and the local loss value, and performing model training on the feature extraction model to be trained according to the total loss value, the method further includes:

[0021] Determine, according to the local feature map and the global momentum feature map, a contrast loss value between the global feature and the local feature corresponding to the feature extraction model to be trained, wherein the magnitude of the contrast loss value is negatively correlated with the similarity between the local feature map and the global momentum feature map;

[0022] Determining a total loss value corresponding to the feature extraction model to be trained according to the global loss value and the local loss value, specifically comprising:

[0023] Determine a total loss value corresponding to the feature extraction model to be trained according to the global loss value, the local loss value and the contrast loss value.

[0024] Optionally, the geomagnetic feature map is input into a preset scale division model so that the scale division model generates scale feature maps corresponding to the geomagnetic information map according to the geomagnetic feature map, specifically including:

[0025] The geomagnetic feature map is input into the scale division model, so that the scale division model generates each scale feature map corresponding to the geomagnetic feature map through each feature extraction layer preset in the scale division model according to the geomagnetic feature map, wherein, for each feature extraction layer, the scale feature map corresponding to the scale feature map output by the previous feature extraction layer of the feature extraction layer is input into the feature extraction layer, so that the feature extraction layer generates the scale feature map corresponding to the feature extraction layer according to the scale feature map corresponding to the previous feature extraction layer, based on the feature extraction resolution and feature receptive field preset for the feature extraction layer.

[0026] Optionally, for each scale feature map, determining a candidate geomagnetic adaptation region in the target region at an image resolution corresponding to the scale feature map specifically includes:

[0027] For each scale feature map, the scale feature map is input into a preset regional prediction model, so that the regional prediction model determines at least part of the target pixels from the scale feature map according to the feature data corresponding to each pixel in the scale feature map, and determines the selected geomagnetic adaptation area in the target area at the image resolution corresponding to the scale feature map according to each target pixel and the corresponding positioning information of each target pixel in the scale feature map, and determines the confidence of the selected geomagnetic adaptation area corresponding to the scale feature map according to the feature data corresponding to each target pixel.

[0028] Optionally, determining the geomagnetic adaptation area corresponding to the target area according to the selected geomagnetic adaptation areas corresponding to the scale feature maps specifically includes:

[0029] According to the confidences of the candidate geomagnetic adaptation regions corresponding to the scale feature maps, the candidate geomagnetic adaptation region with the greatest confidence is taken as the target candidate geomagnetic adaptation region;

[0030] For each candidate geomagnetic adaptation region other than the target candidate geomagnetic adaptation region, determine the intersection-and-union ratio between the target candidate geomagnetic adaptation region and the candidate geomagnetic adaptation region, and when the intersection-and-union ratio is greater than a preset intersection-and-union ratio threshold, eliminate the candidate geomagnetic adaptation region, and when the intersection-and-union ratio is not greater than the preset intersection-and-union ratio threshold, retain the candidate geomagnetic adaptation region;

[0031] The target candidate geomagnetic adaptation area is re-determined from the retained candidate geomagnetic adaptation areas, and the intersection and union ratio of the re-determined target candidate geomagnetic adaptation area and each of the remaining retained candidate geomagnetic adaptation areas is determined until no candidate geomagnetic adaptation areas are retained, so as to determine the geomagnetic adaptation area corresponding to the target area.

[0032] This specification provides a geomagnetic adaptation area positioning device for geomagnetic navigation, including:

[0033] An acquisition module is used to acquire a geomagnetic information map corresponding to a target area, wherein the geomagnetic information map is used to characterize the geomagnetic distribution in the target area. For each grid point preset in the target area, the geomagnetic information map records geomagnetic data corresponding to the grid point, and the geomagnetic data corresponding to the grid point at least includes three-dimensional geomagnetic vector data corresponding to the grid point or combined data composed of the total geomagnetic field strength, magnetic declination and magnetic inclination corresponding to the grid point;

[0034] A feature determination module, used for determining a geomagnetic feature map corresponding to the target area according to the geomagnetic information map, and inputting the geomagnetic feature map into a preset scale division model, so that the scale division model generates each scale feature map corresponding to the geomagnetic information map according to the geomagnetic feature map, and for each scale feature map, the scale feature map is used to represent the geomagnetic feature map of the geomagnetic information map at the image resolution corresponding to the scale feature map;

[0035] A first determination module is used to determine, for each scale feature map, a candidate geomagnetic adaptation area in the target area at an image resolution corresponding to the scale feature map;

[0036] The second determination module is used to determine the geomagnetic adaptation area corresponding to the target area according to the selected geomagnetic adaptation areas corresponding to the scale feature maps.

[0037] The present specification provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned geomagnetic adaptation area positioning method for geomagnetic navigation is implemented.

[0038] This specification provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the above-mentioned geomagnetic adaptation area positioning method for geomagnetic navigation is implemented.

[0039] At least one of the above technical solutions adopted in this specification can achieve the following beneficial effects:

[0040] It can be seen from the above method that in the geomagnetic adaptation area positioning method for geomagnetic navigation provided in this specification, the geomagnetic feature map corresponding to the target area is determined according to the geomagnetic information map corresponding to the target area, and the geomagnetic feature map is input into the preset scale division model, and the geomagnetic feature map corresponding to the geomagnetic information map at each image resolution is generated by the scale division model. Then, for each scale feature map, the selected geomagnetic adaptation area in the target area at the corresponding image resolution is determined. According to each selected geomagnetic adaptation area, the geomagnetic adaptation area in the target area is determined.

[0041] From the above content, it can be seen that the geomagnetic adaptation area positioning method for geomagnetic navigation provided in this specification can determine the corresponding geomagnetic feature map according to the geomagnetic information map corresponding to the target area, and based on the geomagnetic feature map, predict the geomagnetic adaptation area in the target area that can be used for track planning. Compared with the current technology of manually observing and analyzing the geomagnetic information map to determine the geomagnetic adaptation area of ​​the target area, the method in this specification is more intelligent and has a higher degree of automation. While the efficiency and accuracy of determining the geomagnetic adaptation area are improved, it also greatly saves the consumption of human resources and time costs, so that the subsequent execution efficiency of geomagnetic route planning in the target area based on the geomagnetic adaptation area is also improved to a certain extent. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The drawings described herein are used to provide a further understanding of this specification and constitute a part of this specification. The illustrative embodiments and descriptions of this specification are used to explain this specification and do not constitute an improper limitation on this specification. In the drawings:

[0043] Figure 1 A schematic diagram of a flow chart of a geomagnetic adaptation area positioning method for geomagnetic navigation provided in this specification;

[0044] Figure 2 A flowchart of a model training process of a feature extraction model provided in this specification;

[0045] Figure 3 A schematic diagram of a geomagnetic adaptation area positioning device for geomagnetic navigation provided in this specification;

[0046] Figure 4 A method provided in this specification corresponding to Figure 1 Schematic diagram of the structure of an electronic device. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solutions and advantages of this specification more clear, the technical solutions of this specification will be clearly and completely described below in combination with the specific embodiments of this specification and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of this specification, not all of them. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this specification.

[0048] The technical solutions provided by the embodiments of this specification are described in detail below in conjunction with the accompanying drawings.

[0049] Figure 1 A schematic diagram of a flow chart of a geomagnetic adaptation area positioning method for geomagnetic navigation provided in this specification, comprising the following steps:

[0050] S101: Obtain a geomagnetic information map corresponding to a target area.

[0051] At present, navigation technology has gradually matured and is well-known to everyone. Its practicality has brought great convenience to various technical fields and people's daily lives. Geomagnetic navigation technology is a relatively special type of navigation technology. It has a wider range of application scenarios and does not produce cumulative errors during the application process. It is a beneficial supplement to other navigation technologies, such as inertial navigation. In geomagnetic navigation technology, it is necessary to predetermine the geomagnetic adaptation area that can be used for track planning. A reasonably set geomagnetic adaptation area can provide better positioning assistance when performing geomagnetic navigation.

[0052] At present, the determination of geomagnetic adaptation areas in related technologies is mostly done by manually observing and analyzing the geomagnetic data collected in the target area, thereby selecting some areas in the target area with large differences in geomagnetic characteristics as geomagnetic adaptation areas. This method is determined by human subjectivity, and it is highly subjective, with a high missed detection rate and low determination efficiency. Therefore, how to efficiently determine the geomagnetic adaptation area in the target area is a very important issue.

[0053] To this end, this specification provides a geomagnetic adaptation area positioning method for geomagnetic navigation, wherein the execution subject adopted by the method provided in this specification can be a terminal device such as a desktop computer, a laptop computer, or a server. In addition, the execution subject of this specification can also be a subject in the form of software, such as a client installed in a terminal device. For the convenience of explanation of the subsequent process, this specification below only uses the terminal device as the execution subject to explain the provided geomagnetic adaptation area positioning method for geomagnetic navigation.

[0054] Based on this, a terminal device using a method for locating a geomagnetic adaptation area for geomagnetic navigation provided in this specification can determine the geomagnetic feature map corresponding to the geomagnetic information map according to the geomagnetic information map corresponding to the target area, and locate the geomagnetic adaptation area from the target area according to the geomagnetic feature map. Among them, the actual application scenario corresponding to the determined geomagnetic adaptation area can be determined according to actual needs, for example: flight trajectory navigation in the aviation field, when aircraft and other flight equipment fly to certain specific areas, they may encounter problems such as loss of satellite positioning that affect normal navigation accuracy. The terminal device can determine the geomagnetic feature map corresponding to the target area according to the geomagnetic information map corresponding to the target area. Then, the terminal device can determine the geomagnetic adaptation area that can be used for aviation flight trajectory planning from the target area according to the geomagnetic feature map. Relevant professionals can set the route based on the determined geomagnetic adaptation area, so that the flight equipment can rely on geomagnetic navigation to continue flying when it flies near the target area.

[0055] For example, in underground operations and underground positioning, since underground mines, tunnels and other scenes are far away from the ground and the satellite signal is weak, ordinary satellite navigation and positioning methods are easily affected and interfered. Geomagnetic navigation technology can be used to determine the relevant information of the target location even at a deep underground location. In such scenarios, the terminal device using the method of this specification can determine the geomagnetic feature map containing the underground area based on the corresponding geomagnetic information map in the target area, and then determine the geomagnetic adaptation area with obvious geomagnetic feature differences in the target area based on the geomagnetic feature map. When performing underground operations, the target object or personnel can be located based on the determined geomagnetic adaptation area to achieve accurate and real-time grasp of the location information, so that the underground operation tasks can improve the execution efficiency.

[0056] In this specification, the terminal device can obtain a geomagnetic information map corresponding to the target area. The geomagnetic information map can be used to characterize the geomagnetic distribution in the target area. Specifically, it can be in the form of a geomagnetic heat map. For example, the RGB color corresponding to each pixel in the geomagnetic heat map represents the three-dimensional geomagnetic vector distribution of the corresponding preset grid points in the target area and other data. The grid points can be determined in advance based on the minimum unit area in the target area according to a preset division rule.

[0057] What needs to be explained about the above content is that the present specification does not limit the specific means or methods for obtaining the geomagnetic characteristic map corresponding to the target area mentioned above. For example, the geomagnetic data in the target area can be obtained by collecting equipment for geomagnetic data. The collection equipment for geomagnetic data can be a carrier with autonomous movement capabilities such as an unmanned vehicle, an unmanned aerial vehicle, an unmanned boat, etc. The geomagnetic data in the target area is collected by geomagnetic data collection devices such as vector geomagnetic sensors carried on the collection equipment. The collection equipment can also perform data preprocessing on the directly collected geomagnetic data, such as data correction, outlier filtering, etc., and generate a geomagnetic information map corresponding to the target area based on the processed collected geomagnetic data.

[0058] In addition, the geomagnetic information map obtained above may include geomagnetic data corresponding to each grid point in the target area, and the geomagnetic data corresponding to each grid point may at least include three-dimensional geomagnetic vector data corresponding to the grid point or any one of the combined data. The combined data corresponding to the grid point may at least include relevant geomagnetic data such as the total geomagnetic field strength, magnetic declination and magnetic inclination corresponding to the grid point, or it may also be other forms of physical parameters corresponding to the grid point that can comprehensively describe the corresponding geomagnetic field characteristics. The specific data content and type can also be adjusted according to the actual application scenario and the geographical location of the target area, without strict limitation.

[0059] The specific processing procedure of how to generate a geomagnetic information map based on the collected geomagnetic data is also not strictly limited in this specification. For example, various types of geomagnetic data can be first converted into the data form of a multi-channel two-dimensional array, and then the geomagnetic information map corresponding to the target area can be determined based on the two-dimensional geomagnetic data corresponding to each grid point in the target area.

[0060] S102: Determine a geomagnetic feature map corresponding to the target area according to the geomagnetic information map, and input the geomagnetic feature map into a preset scale division model, so that the scale division model generates scale feature maps corresponding to the geomagnetic information map according to the geomagnetic feature map.

[0061] In this specification, the terminal device can determine the geomagnetic feature map corresponding to the target area based on the acquired geomagnetic information map. Then, the terminal device can input the determined geomagnetic feature map into the preset scale division model, so that the scale division model can generate each scale feature map corresponding to the geomagnetic information map according to the received geomagnetic feature map. Among them, each scale feature map corresponds to an image resolution, which can be specifically understood as the feature map representation of the geomagnetic information map at each scale feature map corresponding to the image resolution.

[0062] Regarding the specific processing process of how to determine the corresponding geomagnetic feature map based on the geomagnetic information map, the geomagnetic information map that can be obtained by the terminal device is input into a pre-trained feature extraction model, so that the feature extraction model can extract the geomagnetic feature map corresponding to the target area based on the received geomagnetic information map.

[0063] Regarding the pre-training process of the feature extraction model, in this specification, the terminal device may perform pre-model training on the feature extraction model before using the feature extraction model for actual application to ensure the feature extraction capability of the feature extraction model. The model training method of the feature extraction model in this specification may choose to adopt a self-supervised learning training method, that is, it does not rely on label information, but guides the model to learn the intrinsic characteristics of the input data by designing a specific loss function.

[0064] Specifically, the terminal device can obtain the sample geomagnetic information map used for the model training process, and perform global image enhancement processing and local image enhancement processing on the sample geomagnetic information map to obtain a global enhanced image and a local enhanced image corresponding to the sample geomagnetic information map.

[0065] The specific process of global image enhancement processing may include at least one enhancement processing such as image size cropping, rotation processing and flipping processing for the sample geomagnetic information map, so as to obtain a global enhanced image corresponding to the sample geomagnetic information map. The above rotation processing may be a clockwise or counterclockwise rotation with the axis perpendicular to the image plane corresponding to the center point of the sample geomagnetic information map as the center axis, and the flipping processing may be a mirror flipping of the sample geomagnetic information map up and down or left and right.

[0066] The local image enhancement process is similar to the global image enhancement process, except that the local image enhancement process is performed on the entire image according to the global image enhancement process described above, and then the sample geomagnetic information map is segmented according to a preset segmentation ratio, for example, the complete sample geomagnetic information map is segmented into 16 sample local images of equal size according to a 4*4 segmentation ratio. Then, all the sample local geomagnetic information maps that have undergone image enhancement are spliced ​​into a geomagnetic information image according to their original positions and adjacent relationships as the local enhanced image corresponding to the sample geomagnetic information map.

[0067] Alternatively, the terminal device may also perform segmentation and then enhancement, that is, the sample geomagnetic information map is first divided into equal parts according to a preset segmentation ratio, and then random image enhancement processing is performed on each segmented sample local image, and each sample local image after image enhancement is spliced ​​into a geomagnetic information map according to the original position and adjacent relationship as the local enhanced image corresponding to the sample geomagnetic information map. The specific operation sequence and processing process of the local image enhancement processing are not strictly limited in this specification and can be adjusted according to actual application scenarios and needs.

[0068] Regarding the above-mentioned local image enhancement processing process for the geomagnetic information map, specifically, the terminal device can first randomly crop the original entire geomagnetic information map according to a certain image ratio (such as 60% of the entire geomagnetic information map). Then, the terminal device can enlarge the partial sample geomagnetic information map obtained after random cropping, so that the partial sample geomagnetic information map obtained after random cropping is enlarged to the size of the original entire geomagnetic information map, and the pixel ratio of the enlarged partial sample geomagnetic information map is set to m×m, where m is a preset pixel value. Next, the terminal device can perform a similar process to the above-mentioned global image enhancement processing on the cropped and enlarged partial sample geomagnetic information map, that is, perform image enhancement processing on the partial sample geomagnetic information map.

[0069] Furthermore, the terminal device can perform segmentation processing on the enhanced part of the sample geomagnetic information map according to a preset segmentation ratio, and divide the part of the sample geomagnetic information map into n×n blocks, where n is the preset segmentation ratio, and the height and width pixels of each block are [m / n] rounded and added 1. Then, the terminal device can also perform the above-mentioned random cropping processing on each block according to the preset cropping ratio (such as 80%). The cropping process here can be used to avoid overlapping areas between the segmented blocks. Finally, the terminal device can use the geomagnetic information image spliced ​​from the cropped segmented blocks according to the original image position as the local enhanced image corresponding to the sample geomagnetic information map. Regarding the other local enhancement method of segmentation first and then enhancement mentioned above, the processing process is similar to the above, only the processing order changes, and it will not be repeated here.

[0070] Next, the terminal device can simultaneously input the global enhanced image and the local enhanced image corresponding to the sample geomagnetic information map into the feature extraction model to be trained, so that the feature extraction model to be trained can generate a global feature map and a global momentum feature map corresponding to the sample geomagnetic information map according to the global enhanced image corresponding to the sample geomagnetic information map, and at the same time generate a local feature map and a local momentum feature map corresponding to the sample geomagnetic information map according to the local enhanced image corresponding to the sample geomagnetic information map.

[0071] Then, the terminal device can determine the global loss value corresponding to the feature extraction model to be trained based on the global feature map and the global momentum feature map generated by the feature extraction model to be trained, and determine the local loss value corresponding to the feature extraction model to be trained based on the local feature map and the local momentum feature map. Further, based on the global loss value and the local loss value, the terminal device can determine the total loss value corresponding to the feature extraction model to be trained, and adjust and optimize the parameters of the feature extraction model to be trained with the goal of reducing the total loss value, so as to realize the model training process corresponding to the feature extraction model to be trained. Among them, the size of the global loss value is negatively correlated with the similarity between the global feature map and the global momentum feature map, and the size of the local loss value is also negatively correlated with the similarity between the local feature map and the local momentum feature map.

[0072] It should be noted that in the process of determining the above-mentioned feature maps, the terminal device is mainly implemented by the global encoder and global momentum encoder preset in the feature extraction model, as well as the local encoder and local momentum encoder. In this specification, a global momentum encoder for the global encoder and a local momentum encoder for the local encoder can be preset in the feature extraction model. The global momentum encoder and the local momentum encoder update the parameters in the form of momentum update during the model training process. Specifically, it can be understood that the model parameters corresponding to the momentum encoder can be gradually approached according to the parameters of the corresponding encoder in a certain proportion during the model optimization process, so that the parameter optimization and update of the corresponding encoder are more stable, and the model parameters are prevented from falling into the local minimum.

[0073] Specifically, after the terminal device inputs the global enhanced image and the local enhanced image into the feature extraction model to be trained, the feature extraction model can determine the corresponding global feature map based on the global enhanced image through a preset global encoder, and determine the corresponding global momentum feature map through a preset global momentum encoder, and can determine the corresponding local feature map based on the local enhanced image through a preset local encoder, and determine the corresponding local momentum feature map through a preset local momentum encoder.

[0074] Then, the terminal device can determine the global loss value based on the determined global feature map and global momentum feature map. For the specific process of determining the global loss value, refer to the following formula:

[0075]

[0076] in, Specifically used to represent the feature extraction model The global loss value corresponding to the layer, the right side of the equation It is used to represent the first The global feature map extracted by the layer model structure, It is used to represent the sample geomagnetic information map in the The global momentum feature map corresponding to the layer is used as the positive sample corresponding to the global enhanced image of the sample geomagnetic information map. It is used to represent the first Tier global momentum feature map samples as negative samples corresponding to the global enhanced images of the sample geomagnetic information map, is the preset hyperparameter, They are used to represent the global encoder and global momentum encoder in the feature extraction model respectively.

[0077] By calculating the above formula, the terminal device can determine the global loss value corresponding to the feature extraction model to be trained based on the global feature map and the global momentum feature map, so that the global feature map output by the global encoder Closer to positive sample data , and stay away from negative sample data While effectively avoiding the overfitting problem of the global encoder, it can also improve the global encoder's ability to resolve geomagnetic information maps of different scales and improve the accuracy of subsequent geomagnetic adaptation area predictions.

[0078] As for the specific determination process corresponding to the above local loss value, please refer to the following formula:

[0079]

[0080] in, Specifically used to represent the feature extraction model The local loss value corresponding to the layer, the right side of the equation It is used to represent the local encoder in the feature extraction model. The local feature map extracted by the layer model structure, It is used to represent the sample geomagnetic information map in the The local momentum feature map corresponding to the layer is used as the positive sample corresponding to the local enhanced image of the sample geomagnetic information map. It is used for the first local momentum feature map samples as negative samples corresponding to the local enhanced images of the sample geomagnetic information map, is the preset hyperparameter, They are used to represent the local encoder and local momentum encoder in the feature extraction model respectively. Through the above formula calculation, the terminal device can determine the local loss value corresponding to the feature extraction model to be trained based on the local feature map and the local momentum feature map, so that the local feature map output by the local encoder Closer to positive sample data , and stay away from negative sample data .

[0081] In addition, considering that there may be certain connections and similarities between the feature representations corresponding to the global enhanced image and the local enhanced image, in this specification, the terminal device can also determine the contrast loss value between the global features and the local features corresponding to the feature extraction model based on the global momentum feature map and the local feature map corresponding to the sample geomagnetic information map. The specific determination process of the contrast loss value can refer to the following formula:

[0082]

[0083] in, Specifically used to represent the feature extraction model The global and local contrast loss values ​​corresponding to the layer, in the above formula Specifically used to represent the first The local feature map extracted by the layer model structure. It is used to represent the sample geomagnetic information map in the The global momentum feature map corresponding to the layer is used as the positive sample corresponding to the local enhanced image of the sample geomagnetic information map. It is used to represent the first global momentum feature map samples as negative samples corresponding to the local enhanced images of the sample geomagnetic information map, is the preset hyperparameter.

[0084] Through the above formula calculation, the terminal device can determine the contrast loss value corresponding to the feature extraction model to be trained based on the local feature map and the global momentum feature map, so that when the feature extraction model is optimized according to the loss value during the model training process, it can simultaneously learn the relationship between the global and local geomagnetic features, thereby enhancing the model capability and making the geomagnetic adaptation area determined in practical applications more accurate.

[0085] Furthermore, after determining the global loss value, local loss value, and contrast loss value corresponding to the feature extraction model through the above method, the terminal device can determine the total loss value corresponding to the feature extraction model to be trained based on the above multiple loss values. The specific determination process of the total loss value can refer to the following formula:

[0086]

[0087] in, Specifically used to represent the total loss value corresponding to the feature extraction model, The above feature extraction model is The global loss value corresponding to the layer, For the above feature extraction model The local loss value corresponding to the layer, and It means that the feature extraction model The contrast loss value corresponding to the layer, is the preset weight parameter of each feature layer. Through the calculation of this formula, the terminal device can determine the total loss value corresponding to the feature extraction model to be trained, and then optimize the model and adjust the parameters of the feature extraction model with the goal of reducing the total loss value to obtain a feature extraction model with the ability to extract geomagnetic features.

[0088] Regarding the specific processing process of momentum update mentioned in the above process, in this specification, whether it is for the global encoder and the global momentum encoder, or the local encoder and the local momentum encoder, the parameter update can be realized by referring to the following formula:

[0089]

[0090] in, is the preset parameter update ratio corresponding to the momentum encoder in the feature extraction model, is the model parameter of the momentum encoder at the current moment, is the model parameter of the momentum encoder at the previous moment, It is used to represent the model parameters of the corresponding encoder of the momentum encoder at the current moment. Through the above calculation formula, the momentum encoder in the feature extraction model can adjust the momentum update of its own parameter changes according to the parameter changes of the corresponding encoder, thereby making the model training and optimization process of the overall feature extraction model more stable, and the actual application capability of the feature extraction model finally trained is more practical and accurate.

[0091] In addition to the above, in order to improve the model training effect of the feature extraction model, in this specification, when the feature extraction model to be trained is trained, two different global enhanced images corresponding to the sample geomagnetic information map can be generated by image enhancement processing, and two different local enhanced images corresponding to the sample geomagnetic information map can be generated by image enhancement processing for the local geomagnetic image. The terminal device can input two global enhanced images with different enhancement effects into the global encoder and global momentum encoder preset in the feature extraction model to be trained, and input two local enhanced images with different enhancement effects into the local encoder and local momentum encoder preset in the feature extraction model to be trained, and then perform the above training process. By performing different enhancement processing on the same sample geomagnetic information map and then inputting it into the feature extraction model at the same time to extract geomagnetic features, the feature extraction capability of the model for different geomagnetic information maps can be effectively improved, thereby significantly improving the accuracy of the final geomagnetic adaptation area.

[0092] In order to comprehensively understand the model training process for the feature extraction model, the following is a flowchart of the model training process of a feature extraction model to explain it. Figure 2 shown.

[0093] Figure 2 A flowchart of a model training process of a feature extraction model provided in this specification.

[0094] like Figure 2 As shown in , the terminal device can perform image enhancement processing on the sample geomagnetic information map to obtain two different global enhanced images corresponding to the sample geomagnetic information map ( Figure 2 As shown in and ) and two different local enhanced images ( Figure 2 As shown in and ). Then, the terminal device can input the global enhanced image into the global encoder and the global momentum encoder in the feature extraction model to be trained, respectively, so that the global encoder determines the global feature map corresponding to the sample geomagnetic information map based on the received global enhanced image, and the global momentum encoder determines the global momentum feature map corresponding to the sample geomagnetic information map based on the received global enhanced image.

[0095] At the same time, the terminal device can also input the local enhanced image into the local encoder and local momentum encoder in the feature extraction model to be trained, so that the local encoder can determine the local feature map corresponding to the sample geomagnetic information map based on the received local enhanced image, and the local momentum encoder can determine the local momentum feature map corresponding to the sample geomagnetic information map based on the received local enhanced image.

[0096] Next, the terminal device can determine the feature extraction model to be trained based on the global feature map and the global momentum feature map. The global loss value corresponding to the layer , and according to the local feature map and the local momentum feature map, determine the feature extraction model to be trained The local loss value corresponding to the layer At the same time, the terminal device can also determine the contrast loss value between the global features and the local features corresponding to the i-th layer of the feature extraction model to be trained based on the local feature map and the global momentum feature map. Then, according to the loss values ​​of each layer, the total loss value is determined , perform model training and optimization on the feature extraction model.

[0097] It should also be noted that in the above-mentioned training process for the feature extraction model, the geomagnetic features of different enhanced images of the sample geomagnetic information map are extracted mainly through various types of encoders preset in the feature extraction model, so as to ensure that the finally trained model has the ability to accurately extract geomagnetic features. In practical applications, there is no need to perform image enhancement processing on the geomagnetic information map corresponding to the target area, and when the geomagnetic information map corresponding to the target area is directly input into the trained feature extraction model, the feature extraction model can directly determine the global geomagnetic feature map corresponding to the geomagnetic information map through the trained global encoder, and use it as the geomagnetic feature map corresponding to the target area.

[0098] In addition to the self-supervised learning training method for the feature extraction model mentioned above, in this specification, the feature extraction model can also be trained and optimized through other training methods. For example, the feature extraction model to be trained can also be trained through supervised learning, semi-supervised learning, reinforcement learning and other training methods.

[0099] Taking the supervised learning method as an example, specifically, the terminal device can input the acquired sample geomagnetic information map into the feature extraction model to be trained, so that the feature extraction model to be trained performs feature extraction on the received sample geomagnetic information map to obtain the geomagnetic feature map corresponding to the sample geomagnetic information map.

[0100] Then, the terminal device can determine the training loss value corresponding to the feature extraction model according to the geomagnetic feature map generated by the feature extraction model to be trained and the standard geomagnetic feature map corresponding to the sample geomagnetic information map. The size of the training loss value is negatively correlated with the similarity between the geomagnetic feature map generated by the trained feature extraction model and the standard geomagnetic feature map corresponding to the sample geomagnetic information map. The more similar the geomagnetic feature map generated by the trained feature extraction model is to the standard geomagnetic feature map, the smaller the training loss value corresponding to the feature extraction model is, indicating that the model has a stronger ability to extract geomagnetic features.

[0101] The momentum encoder in the above-mentioned self-supervised learning method can also be used in supervised learning. Through the special update method of the momentum encoder based on the main encoder to update the momentum, when determining the training loss and optimizing the model for the encoder in the feature extraction model, the momentum encoder can be used as an auxiliary update network in the feature extraction model. It can slowly update the parameters through its own momentum finer mechanism and record the historical parameter changes of the encoder, thereby providing a more stable geomagnetic feature representation, helping the feature extraction model to converge faster during training and improve the robustness of the model, thereby effectively avoiding the occurrence of overfitting problems.

[0102] In summary of the above training process of the feature extraction model, in this specification, the feature extraction model can be trained through the self-supervised training method and the momentum update method of the momentum encoder, so as to obtain a feature extraction model that can extract the global feature map according to the geomagnetic information map. The training method of self-supervised learning and the establishment of global and local momentum encoders can effectively improve the feature extraction ability of the feature extraction model, and the momentum update method based on the momentum encoder can effectively ensure the stability of the parameter update of the feature extraction model during the training process, accelerate the convergence of the model parameters, and effectively avoid the overfitting problem of the trained model.

[0103] In addition, in this specification, the feature extraction model can also be trained by other types of training methods, such as the supervised learning method mentioned above. Although the training annotation samples are limited and the generalization is slightly weaker than the training process of the above-mentioned self-supervised learning, the model loss is determined according to the geomagnetic feature map generated by the feature extraction model to be trained and the standard geomagnetic feature map corresponding to the sample geomagnetic information map, so that the model can be optimized and trained to obtain a feature extraction model with the ability to extract the geomagnetic feature map according to the geomagnetic information map. Therefore, no matter which training method is used to obtain the trained feature extraction model, it can be used to determine the geomagnetic feature map corresponding to the target area according to the geomagnetic information map corresponding to the target area, and then the terminal device can perform the prediction process of the geomagnetic adaptation area in the subsequent target area based on the determined geomagnetic feature map.

[0104] It should also be noted that in this specification, the specific model type of the feature extraction model mentioned in the above method is not strictly limited. It can be a model structure such as a residual neural network (Residual Network, ResNet), a Darknet model, and a combination with a multilayer perceptron (Multilayer Perceptron, MLP), which can be flexibly set according to actual application scenarios and needs.

[0105] Further, the feature extraction model is obtained by training through the above method, and the geomagnetic information map corresponding to the target area is input into the feature extraction model. After determining the geomagnetic feature map corresponding to the target area, the terminal device can then input the geomagnetic feature map into the preset scale division model, so that the scale division model can generate each scale feature map corresponding to the geomagnetic information map according to the received geomagnetic feature map. Among them, each generated scale feature map corresponds to the feature acquisition resolution of a certain hierarchical structure in the size division model, which can be specifically expressed as the geomagnetic feature map corresponding to the image resolution of the geomagnetic information map at the image resolution corresponding to the scale feature map. The image resolution corresponding to the scale feature map can be obtained by special processing of the image resolution corresponding to the geomagnetic feature map, which can be the same as the feature extraction resolution corresponding to the hierarchical structure of the scale feature map generated in the size division model. The special processing can be specifically scaling the image resolution corresponding to the geomagnetic feature map according to the preset scaling scale to obtain the image resolution corresponding to different scaling scales, or other processing operations for the image resolution, which are not strictly limited in this specification and can be flexibly set according to the actual application scenario.

[0106] Specifically, the terminal device can input the determined geomagnetic feature map into a preset scale division model, so that the scale division model can generate various scale feature maps corresponding to the geomagnetic feature map according to the received geomagnetic feature map through various feature extraction layers preset in the scale division model.

[0107] Regarding the specific processing process of generating a scale feature map by the feature extraction layer in the scale division model, for each feature extraction layer, the feature extraction layer can determine the scale feature map corresponding to the output of the feature extraction layer based on the scale feature map corresponding to the scale feature map output by the previous feature extraction layer, based on the preset feature extraction resolution and feature receptive field corresponding to the feature extraction layer. The first feature extraction layer can directly determine the scale feature map based on the geomagnetic feature map received by the scale division model.

[0108] The scale division model is preset with a multi-layer and interconnected feature extraction layer structure. The feature extraction resolution and feature receptive field corresponding to each feature extraction layer are different, and can be a progressive relationship. While the feature extraction resolution gradually decreases, the feature receptive field can gradually increase. By generating feature maps of each scale corresponding to the geomagnetic information map through the scale division model, the feature data in the target area at different scales can be effectively captured, thereby improving the accuracy of the subsequent determination of the geomagnetic adaptation area.

[0109] The specific model type of the scale division model is not strictly limited in this specification. For example, model structures such as Feature Pyramid Network (FPN) and Path Aggregation Feature Pyramid Network (PAFPN) can be selected, and can be flexibly selected according to actual application scenarios and needs.

[0110] S103: For each scale feature map, determine a candidate geomagnetic adaptation region in the target area at an image resolution corresponding to the scale feature map.

[0111] In this specification, for each scale feature map determined by the above method, the terminal device can determine the selected geomagnetic adaptation area in the target area at the image resolution corresponding to the scale feature map based on the scale feature map and the geomagnetic information map corresponding to the target area.

[0112] Specifically, for each scale feature map, the terminal device can input the scale feature map into a preset regional prediction model, so that the regional prediction model can determine the target pixel points in the scale feature map that are suitable as the geomagnetic adaptation area and the corresponding positioning information of the target pixel points in the scale feature map based on the scale feature map.

[0113] Then, the terminal device can determine the candidate geomagnetic adaptation area in the target area at the image resolution corresponding to the scale feature map through the regional prediction model according to the positioning information corresponding to each target pixel point and the geomagnetic information map corresponding to the target area.

[0114] In addition, the terminal device can also determine the confidence of the selected geomagnetic adaptation area corresponding to the scale feature map according to the feature data corresponding to each target pixel in the scale feature map. The confidence can be specifically used to indicate the matching degree of the selected geomagnetic adaptation area as the geomagnetic adaptation area in the target area.

[0115] Regarding the specific processing process of determining the selected geomagnetic adaptation area according to the scale feature map by the regional prediction model, the terminal device can determine some or all of the pixels with obvious feature differences in the scale feature map as target pixels according to the feature data corresponding to each pixel in each scale feature map through the regional prediction model. According to the position distribution of each target pixel in the scale feature map, the positioning information corresponding to each target pixel is determined. The terminal device can combine the positioning information corresponding to each target pixel into the positioning information of the selected geomagnetic adaptation area corresponding to the scale feature map, and then determine the selected geomagnetic adaptation area corresponding to the scale feature map from the geomagnetic information map corresponding to the target area according to the positioning information of the selected geomagnetic adaptation area and the scaling scale corresponding to the scale feature map.

[0116] In addition, the terminal device can also determine the confidence corresponding to the selected geomagnetic adaptation area through the regional prediction model according to the similarity between the feature data of each adjacent target pixel point in the selected geomagnetic adaptation area corresponding to the scale feature map. The closer the similarity, the lower the confidence. Conversely, the more obvious the difference between the geomagnetic feature maps of adjacent target pixels, the more suitable the selected geomagnetic adaptation area is as the geomagnetic adaptation area, and the higher the confidence.

[0117] It should be noted that the specific model category of the above-mentioned regional prediction model is not limited in this specification. It can be a neural network model with feature analysis capabilities, such as a support vector machine model (SVM), a random forest (RandomForest), etc., which can be adjusted according to the actual application scenario.

[0118] In addition, regarding the training and optimization process of the regional prediction model, in this specification, the regional prediction model can be pre-trained to improve the regional prediction effect in practical applications. At the same time, during the regional prediction model training process, other models in the above process can be optimized according to the loss optimization situation, thereby achieving the joint adjustment and update of the overall model.

[0119] Specifically, during the training process of the regional prediction model, the terminal device can input each sample scale feature map corresponding to the sample geomagnetic information map into the regional prediction model to be trained, so that the regional prediction model to be trained determines the predicted geomagnetic adaptation area corresponding to each sample scale feature map based on the received sample scale feature maps.

[0120] Then, the terminal device can determine the intersection-and-union ratio between the predicted magnetic adaptation area corresponding to each sample scale feature map and the labeled geomagnetic adaptation area corresponding to the sample geomagnetic information map. According to the intersection-and-union ratio, the terminal device can determine the cost value between the predicted geomagnetic adaptation area corresponding to each sample scale feature map and the labeled geomagnetic adaptation area, and the terminal device can sort the predicted geomagnetic adaptation area according to the size of the cost value. The predicted geomagnetic adaptation areas corresponding to the preset number of sample scale feature maps before the sorting result are used as positive samples output by the regional prediction model to be trained, and the predicted geomagnetic adaptation areas corresponding to the remaining sample scale feature maps are used as negative samples output by the regional prediction model to be trained, and then the regional prediction model is optimized for the positive and negative sample data.

[0121] The specific determination process of the above intersection-and-union ratio can refer to the following formula:

[0122]

[0123] Among them, for each sample scale feature map corresponding to the sample geomagnetic information map, It can be used to represent the intersection-over-combination ratio between the predicted geomagnetic adaptation area corresponding to the sample scale feature map and the labeled geomagnetic adaptation area corresponding to the sample geomagnetic information map. Indicates the geomagnetic adaptation area of ​​the tag. It indicates the predicted geomagnetic adaptation area corresponding to the sample scale feature map. Indicates the geomagnetic adaptation area of ​​the tag Area compatible with predicted geomagnetic The size of the intersection area, Indicates the geomagnetic adaptation area of ​​the tag Area compatible with predicted geomagnetic The size of the union area.

[0124] As for the specific determination process of the above cost value, please refer to the following formula:

[0125]

[0126] Among them, for each sample scale feature map corresponding to the sample geomagnetic information map, It is used to represent the cost value between the predicted geomagnetic adaptation area and the labeled geomagnetic adaptation area corresponding to the sample scale feature map, specifically the classification confidence loss corresponding to the predicted geomagnetic adaptation area and the labeled geomagnetic adaptation area. And the regression loss corresponding to the predicted geomagnetic adaptation area and the labeled geomagnetic adaptation area Determined by weighted summation, is the preset weight parameter. Classification confidence loss Specifically, it can be used to optimize the regional prediction model's ability to accurately determine the confidence level corresponding to the selected geomagnetic adaptation area, while the regression loss Specifically, it can be used to optimize the regional prediction model's ability to accurately determine the positioning information corresponding to the selected geomagnetic adaptation area. The specific determination process of the two losses can refer to the following formula:

[0127]

[0128]

[0129] As shown in the above formula, in this specification, the terminal device can determine the predicted geomagnetic adaptation area corresponding to the scale feature map through the binary cross entropy loss function shown in the first line of the above formula. The geomagnetic matching area of ​​the tag The classification confidence loss between , and the predicted geomagnetic adaptation area determined previously The geomagnetic matching area of ​​the tag The area intersection ratio , determine the regression loss .

[0130] Then, the terminal device can determine the loss value of the positive and negative samples output by the regional prediction model. For details, please refer to the following formula:

[0131]

[0132]

[0133] in, It is used to represent the loss value corresponding to the predicted geomagnetic adaptation area corresponding to the positive samples in the training process of the regional prediction model, that is, the loss value corresponding to the predicted geomagnetic adaptation area corresponding to the preset number of sample scale feature maps before the cost value sorting result. is the preset weight parameter, and the overall calculation process is basically the same as the cost value calculation process. It is used to represent the loss value corresponding to the predicted geomagnetic adaptation area corresponding to each sample scale feature map except the preset number before the negative samples are sorted according to the cost value in the training process of the regional prediction model.

[0134] Then, the terminal device can calculate the positive sample loss based on and negative sample loss , determine the total loss value corresponding to the regional prediction model, which can be specifically referred to the following formula:

[0135] Then the terminal device can determine the total loss value , train and optimize the regional prediction model.

[0136] It should also be noted that in this specification, considering that the overall geomagnetic adaptation regional positioning method is composed of multiple model structures, in order to ensure the effective joint cooperation of each model in actual operation, the terminal device is based on the total loss value determined by the above method. When training the regional prediction model, it can be converted into training the overall combined model, that is, including the pre-trained feature extraction model, scale division model and regional prediction model in the previous step, so as to improve the connection between the various parts of the overall model and optimize the parameters of the previous model, so that the geomagnetic adaptation area corresponding to the target area can be determined more accurately in practical applications.

[0137] S104: Determine the geomagnetic adaptation region corresponding to the target region according to the candidate geomagnetic adaptation regions corresponding to the scale feature maps.

[0138] In this specification, the terminal device can determine the geomagnetic adaptation area corresponding to the target area based on the candidate geomagnetic adaptation areas corresponding to each scale feature map determined by the above method.

[0139] Specifically, the terminal device may select the candidate geomagnetic adaptation region with the greatest confidence as the target candidate geomagnetic adaptation region according to the confidence of the candidate geomagnetic adaptation region corresponding to each scale feature map determined by the above method.

[0140] Then, for each other candidate geomagnetic adaptation area except the target candidate geomagnetic adaptation area, the terminal device can determine the intersection and union ratio between the candidate geomagnetic adaptation area and the target candidate geomagnetic adaptation area. According to the size of the intersection and union ratio, each candidate geomagnetic adaptation area is filtered or retained. When the intersection and union ratio is greater than the preset intersection and union ratio threshold, the terminal device can consider that the candidate geomagnetic adaptation area overlaps too much with the target candidate geomagnetic adaptation area and the confidence is low, which is an invalid geomagnetic adaptation area, and directly filters it. When the union ratio is not greater than the preset intersection and union ratio threshold, the terminal device can consider that although the confidence is low, the candidate geomagnetic adaptation area overlaps less with the target candidate geomagnetic adaptation area, and can be retained.

[0141] After all the candidate geomagnetic adaptation areas except the target candidate geomagnetic adaptation area have been filtered or retained, the target candidate geomagnetic adaptation area is re-determined from the retained candidate geomagnetic adaptation area according to the confidence level, and the re-determined target candidate geomagnetic adaptation area is compared with the intersection and union of the other previously retained candidate geomagnetic adaptation areas, and the filtering or retention process is continued until no candidate geomagnetic adaptation area is retained. Through the above steps, the combined area of ​​the single or multiple candidate geomagnetic adaptation areas determined can be used as the geomagnetic adaptation area corresponding to the target area.

[0142] In addition to the above methods, in this specification, the terminal device can also determine the geomagnetic adaptation area corresponding to the target area according to each candidate geomagnetic adaptation area through calculation methods such as multi-level NMS (Multi-LevelNMS) and soft NMS (Soft NMS). The determined geomagnetic adaptation area can be output to the relevant track planners, so that the relevant personnel can reasonably plan the geomagnetic navigation track in the target area based on the geomagnetic adaptation area corresponding to the target area.

[0143] As can be seen from the above, the geomagnetic adaptation area positioning method for geomagnetic navigation provided in this specification can determine the corresponding geomagnetic feature map according to the geomagnetic information map corresponding to the target area, and based on the geomagnetic feature map, predict the geomagnetic adaptation area that can be used for track planning in the target area. Through the data processing of each model and the mutual data transmission, the high degree of automation and intelligence of the geomagnetic adaptation area is realized. Compared with the current human determination of the geomagnetic adaptation area based on the geomagnetic information map, the method in this specification greatly improves the efficiency and accuracy of determining the geomagnetic adaptation area, making it more reasonable to carry out subsequent geomagnetic track planning for the target area, while also effectively saving human resources and reducing cost consumption and time waste.

[0144] The above are one or more implementation methods of this specification. Based on the same idea, this specification also provides a corresponding geomagnetic adaptation area positioning device for geomagnetic navigation, such as Figure 3 shown.

[0145] Figure 3 A schematic diagram of a geomagnetic adaptation area positioning device for geomagnetic navigation provided in this specification, including:

[0146] An acquisition module 301 is used to acquire a geomagnetic information map corresponding to a target area, wherein the geomagnetic information map is used to characterize the geomagnetic distribution in the target area. For each grid point preset in the target area, the geomagnetic information map records geomagnetic data corresponding to the grid point, and the geomagnetic data corresponding to the grid point at least includes three-dimensional geomagnetic vector data corresponding to the grid point or combined data composed of the total geomagnetic field strength, magnetic declination and magnetic inclination corresponding to the grid point;

[0147] A feature determination module 302 is used to determine a geomagnetic feature map corresponding to the target area according to the geomagnetic information map, and input the geomagnetic feature map into a preset scale division model, so that the scale division model generates each scale feature map corresponding to the geomagnetic information map according to the geomagnetic feature map, and for each scale feature map, the scale feature map is used to represent the geomagnetic feature map of the geomagnetic information map at the image resolution corresponding to the scale feature map;

[0148] A first determination module 303 is used to determine, for each scale feature map, a candidate geomagnetic adaptation region in the target area at an image resolution corresponding to the scale feature map;

[0149] The second determination module 304 is used to determine the geomagnetic adaptation area corresponding to the target area according to the candidate geomagnetic adaptation areas corresponding to the scale feature maps.

[0150] Optionally, the feature determination module 302 is specifically used to input the geomagnetic information map into a pre-trained feature extraction model, so that the feature extraction model determines the geomagnetic feature map corresponding to the target area based on the geomagnetic information map.

[0151] Optionally, the feature determination module 302 is specifically used to obtain a sample geomagnetic information map; perform image enhancement processing on the sample geomagnetic information map to obtain a global enhanced image corresponding to the sample geomagnetic information map, and perform image enhancement processing on a sample local geomagnetic information map corresponding to the sample geomagnetic information map to obtain a local enhanced image corresponding to the sample geomagnetic information map, wherein the image enhancement processing is at least any one of image size cropping processing, rotation processing and flipping processing, and the sample local geomagnetic information map is first subjected to the image enhancement processing and then subjected to the image enhancement processing. The sample geomagnetic information map after the image enhancement processing is obtained by segmenting according to a preset segmentation ratio, and the local enhanced image is obtained by splicing the sample local geomagnetic information maps corresponding to the sample geomagnetic information map according to the original image position; the global enhanced image and the local enhanced image are input into the feature extraction model to be trained, so that the feature extraction model to be trained determines the global feature map corresponding to the sample geomagnetic information map through a preset global encoder according to the global enhanced image, and determines the global feature map corresponding to the sample geomagnetic information map through a preset global momentum encoder. A local momentum feature map, and a local feature map corresponding to the sample geomagnetic information map is determined by a preset local encoder, and a local momentum feature map corresponding to the sample geomagnetic information map is determined by a preset local momentum encoder, wherein the model parameters of the global momentum encoder are momentum updated according to the model parameters of the global encoder according to a preset update ratio, and the model parameters of the local momentum encoder are momentum updated according to the model parameters of the local encoder according to a preset update ratio; according to the global feature map and the global momentum feature map, a global loss value corresponding to the feature extraction model to be trained is determined, and according to the local feature map and the local momentum feature map, a local loss value corresponding to the feature extraction model to be trained is determined, wherein the size of the global loss value is negatively correlated with the similarity between the global feature map and the global momentum feature map, and the size of the local loss value is negatively correlated with the similarity between the local feature map and the local momentum feature map; according to the global loss value and the local loss value, a total loss value corresponding to the feature extraction model to be trained is determined, and according to the total loss value, model training is performed on the feature extraction model to be trained.

[0152] Optionally, the feature determination module 302 is specifically used to determine, according to the local feature map and the global momentum feature map, a contrast loss value between the global feature and the local feature corresponding to the feature extraction model to be trained, wherein the size of the contrast loss value is negatively correlated with the similarity between the local feature map and the global momentum feature map;

[0153] The feature determination module 302 is specifically used to determine the total loss value corresponding to the feature extraction model to be trained according to the global loss value, the local loss value and the contrast loss value.

[0154] Optionally, the feature determination module 302 is specifically used to input the geomagnetic feature map into the scale division model, so that the scale division model generates each scale feature map corresponding to the geomagnetic feature map according to the geomagnetic feature map through each feature extraction layer preset in the scale division model, wherein, for each feature extraction layer, the scale feature map corresponding to the scale feature map output by the previous feature extraction layer of the feature extraction layer is input into the feature extraction layer, so that the feature extraction layer generates the scale feature map corresponding to the feature extraction layer according to the scale feature map corresponding to the previous feature extraction layer, based on the feature extraction resolution and feature receptive field preset for the feature extraction layer.

[0155] Optionally, the first determination module 303 is specifically used to, for each scale feature map, input the scale feature map into a preset regional prediction model, so that the regional prediction model determines at least part of the target pixels from the scale feature map according to the feature data corresponding to each pixel in the scale feature map, and determines the candidate geomagnetic adaptation area in the target area at the image resolution corresponding to the scale feature map according to each target pixel and the corresponding positioning information of each target pixel in the scale feature map, and determines the confidence of the candidate geomagnetic adaptation area corresponding to the scale feature map according to the feature data corresponding to each target pixel.

[0156] Optionally, the second determination module 304 is specifically used to, according to the confidences of the candidate geomagnetic adaptation regions corresponding to the scale feature maps, take the candidate geomagnetic adaptation region with the largest confidence as the target candidate geomagnetic adaptation region; for each candidate geomagnetic adaptation region other than the target candidate geomagnetic adaptation region, determine the intersection-and-union ratio between the target candidate geomagnetic adaptation region and the candidate geomagnetic adaptation region, and when the intersection-and-union ratio is greater than a preset intersection-and-union ratio threshold, eliminate the candidate geomagnetic adaptation region; when the intersection-and-union ratio is not greater than the preset intersection-and-union ratio threshold, retain the candidate geomagnetic adaptation region; redetermine the target candidate geomagnetic adaptation region from the retained candidate geomagnetic adaptation regions, and determine the intersection-and-union ratio corresponding to the redetermined target candidate geomagnetic adaptation region and each of the remaining retained candidate geomagnetic adaptation regions, until no candidate geomagnetic adaptation region is retained, so as to determine the geomagnetic adaptation region corresponding to the target region.

[0157] This specification also provides a computer-readable storage medium, which stores a computer program, which can be used to execute the above Figure 1 A geomagnetic adaptation regional positioning method for geomagnetic navigation is provided.

[0158] This manual also provides Figure 4 The one shown corresponds to Figure 1 A schematic diagram of the electronic device. Figure 4 As shown, at the hardware level, the electronic device includes a processor, an internal bus, a network interface, a memory, and a non-volatile memory, and may also include other hardware required for the business. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it to achieve the above Figure 1 The geomagnetic adaptation area positioning method for geomagnetic navigation.

[0159] In the 1990s, it was very clear whether the improvement of a technology was hardware improvement (for example, improvement of the circuit structure of diodes, transistors, switches, etc.) or software improvement (improvement of the method flow). However, with the development of technology, many improvements of the method flow today can be regarded as direct improvements of the hardware circuit structure. Designers almost always obtain the corresponding hardware circuit structure by programming the improved method flow into the hardware circuit. Therefore, it cannot be said that the improvement of a method flow cannot be implemented with hardware entity modules. For example, a programmable logic device (PLD) (such as a field programmable gate array (FPGA)) is such an integrated circuit whose logical function is determined by the user's programming of the device. Designers can "integrate" a digital system on a PLD by programming it themselves, without having to ask chip manufacturers to design and make dedicated integrated circuit chips. Moreover, nowadays, instead of manually making integrated circuit chips, this kind of programming is mostly implemented by "logic compiler" software, which is similar to the software compiler used when developing and writing programs, and the original code before compilation must also be written in a specific programming language, which is called hardware description language (HDL). There is not only one kind of HDL, but many kinds, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, RHDL (Ruby Hardware Description Language), etc. The most commonly used ones are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should also know that it is only necessary to program the method flow slightly in the above-mentioned hardware description languages ​​and program it into the integrated circuit, and then it is easy to obtain the hardware circuit that implements the logic method flow.

[0160] The controller may be implemented in any suitable manner, for example, the controller may take the form of a microprocessor or processor and a computer-readable medium storing a computer-readable program code (e.g., software or firmware) executable by the (micro)processor, a logic gate, a switch, an application-specific integrated circuit (ASIC), a programmable logic controller, and an embedded microcontroller, examples of which include but are not limited to the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320, and the memory controller may also be implemented as part of the control logic of the memory. It is also known to those skilled in the art that, in addition to implementing the controller in a purely computer-readable program code manner, the controller may be implemented in the form of a logic gate, a switch, an application-specific integrated circuit, a programmable logic controller, and an embedded microcontroller by logically programming the method steps. Therefore, such a controller may be considered as a hardware component, and the devices for implementing various functions included therein may also be considered as structures within the hardware component. Or even, the devices for implementing various functions may be considered as both software modules for implementing the method and structures within the hardware component.

[0161] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0162] For the convenience of description, the above device is described in various units according to their functions. Of course, when implementing this specification, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0163] Those skilled in the art will appreciate that the embodiments of this specification may be provided as methods, systems, or computer program products. Therefore, this specification may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0164] This specification is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of this specification. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0165] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0166] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0167] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0168] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0169] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0170] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0171] It should be understood by those skilled in the art that the embodiments of this specification may be provided as methods, systems or computer program products. Therefore, this specification may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0172] This specification may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. This specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0173] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0174] The above description is only an embodiment of the present specification and is not intended to limit the present specification. For those skilled in the art, the present specification may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present specification shall be included in the scope of the claims of the present specification.

Claims

1. A geomagnetic adaptation regional positioning method for geomagnetic navigation, characterized in that: include: Obtaining a geomagnetic information map corresponding to the target area, the geomagnetic information map is used to characterize the geomagnetic distribution in the target area, and for each grid point preset in the target area, the geomagnetic information map records geomagnetic data corresponding to the grid point, and the geomagnetic data corresponding to the grid point at least includes three-dimensional geomagnetic vector data corresponding to the grid point or combined data composed of the total geomagnetic field strength, magnetic declination and magnetic inclination corresponding to the grid point; According to the geomagnetic information map, determine the geomagnetic characteristic map corresponding to the target area, and input the geomagnetic characteristic map into a preset scale division model, so that the scale division model generates each scale characteristic map corresponding to the geomagnetic information map according to the geomagnetic characteristic map, and for each scale characteristic map, the scale characteristic map is used to represent the geomagnetic characteristic map of the geomagnetic information map at the image resolution corresponding to the scale characteristic map; For each scale characteristic map, determining a candidate geomagnetic adaptation region in the target area at an image resolution corresponding to the scale characteristic map; According to the selected geomagnetic adaptation areas corresponding to the scale feature maps, the geomagnetic adaptation area corresponding to the target area is determined.

2. The method according to claim 1, characterized in that Determining a geomagnetic characteristic map corresponding to the target area according to the geomagnetic information map specifically includes: The geomagnetic information map is input into a pre-trained feature extraction model, so that the feature extraction model determines the geomagnetic feature map corresponding to the target area based on the geomagnetic information map.

3. The method according to claim 2, characterized in that Train the feature extraction model, including: Get sample geomagnetic information map; Performing image enhancement processing on the sample geomagnetic information map to obtain a global enhanced image corresponding to the sample geomagnetic information map, and performing image enhancement processing on the sample local geomagnetic information map corresponding to the sample geomagnetic information map to obtain a local enhanced image corresponding to the sample geomagnetic information map, wherein the image enhancement processing is at least any one of image size cropping processing, rotation processing and flipping processing, the sample local geomagnetic information map is obtained by first performing the image enhancement processing on the sample geomagnetic information map, and then dividing the sample geomagnetic information map after the image enhancement processing according to a preset dividing ratio, and the local enhanced image is obtained by splicing the sample local geomagnetic information maps corresponding to the sample geomagnetic information map according to the original image position; Input the global enhanced image and the local enhanced image into the feature extraction model to be trained, so that the feature extraction model to be trained determines the global feature map corresponding to the sample geomagnetic information map through a preset global encoder according to the global enhanced image, and determines the global momentum feature map corresponding to the sample geomagnetic information map through a preset global momentum encoder, and determines the local feature map corresponding to the sample geomagnetic information map through a preset local encoder, and determines the local momentum feature map corresponding to the sample geomagnetic information map through a preset local momentum encoder, wherein the model parameters of the global momentum encoder are updated with momentum according to the model parameters of the global encoder at a preset update ratio, and the model parameters of the local momentum encoder are updated with momentum according to the model parameters of the local encoder at a preset update ratio; Determine a global loss value corresponding to the feature extraction model to be trained according to the global feature map and the global momentum feature map, and determine a local loss value corresponding to the feature extraction model to be trained according to the local feature map and the local momentum feature map, wherein the size of the global loss value is negatively correlated with the similarity between the global feature map and the global momentum feature map, and the size of the local loss value is negatively correlated with the similarity between the local feature map and the local momentum feature map; According to the global loss value and the local loss value, a total loss value corresponding to the feature extraction model to be trained is determined, and according to the total loss value, model training is performed on the feature extraction model to be trained.

4. The method according to claim 3, characterized in that Before determining the total loss value corresponding to the feature extraction model to be trained according to the global loss value and the local loss value, the method further includes: Determine, according to the local feature map and the global momentum feature map, a contrast loss value between the global feature and the local feature corresponding to the feature extraction model to be trained, wherein the magnitude of the contrast loss value is negatively correlated with the similarity between the local feature map and the global momentum feature map; Determining a total loss value corresponding to the feature extraction model to be trained according to the global loss value and the local loss value, specifically comprising: Determine a total loss value corresponding to the feature extraction model to be trained according to the global loss value, the local loss value and the contrast loss value.

5. The method according to claim 1, characterized in that Inputting the geomagnetic characteristic map into a preset scale division model so that the scale division model generates each scale characteristic map corresponding to the geomagnetic information map according to the geomagnetic characteristic map, specifically comprising: The geomagnetic feature map is input into the scale division model, so that the scale division model generates each scale feature map corresponding to the geomagnetic feature map through each feature extraction layer preset in the scale division model according to the geomagnetic feature map, wherein, for each feature extraction layer, the scale feature map corresponding to the scale feature map output by the previous feature extraction layer of the feature extraction layer is input into the feature extraction layer, so that the feature extraction layer generates the scale feature map corresponding to the feature extraction layer according to the scale feature map corresponding to the previous feature extraction layer, based on the feature extraction resolution and feature receptive field preset for the feature extraction layer.

6. The method according to claim 1, characterized in that For each scale feature map, determining a candidate geomagnetic adaptation region in the target area at an image resolution corresponding to the scale feature map specifically includes: For each scale feature map, the scale feature map is input into a preset regional prediction model, so that the regional prediction model determines at least part of the target pixels from the scale feature map according to the feature data corresponding to each pixel in the scale feature map, and determines the selected geomagnetic adaptation area in the target area at the image resolution corresponding to the scale feature map according to each target pixel and the corresponding positioning information of each target pixel in the scale feature map, and determines the confidence of the selected geomagnetic adaptation area corresponding to the scale feature map according to the feature data corresponding to each target pixel.

7. The method according to claim 6, characterized in that Determining the geomagnetic adaptation area corresponding to the target area according to the selected geomagnetic adaptation areas corresponding to the scale feature maps specifically includes: According to the confidences of the candidate geomagnetic adaptation regions corresponding to the scale feature maps, the candidate geomagnetic adaptation region with the greatest confidence is taken as the target candidate geomagnetic adaptation region; For each candidate geomagnetic adaptation region other than the target candidate geomagnetic adaptation region, determine the intersection-and-union ratio between the target candidate geomagnetic adaptation region and the candidate geomagnetic adaptation region, and when the intersection-and-union ratio is greater than a preset intersection-and-union ratio threshold, eliminate the candidate geomagnetic adaptation region, and when the intersection-and-union ratio is not greater than the preset intersection-and-union ratio threshold, retain the candidate geomagnetic adaptation region; The target candidate geomagnetic adaptation area is re-determined from the retained candidate geomagnetic adaptation areas, and the intersection and union ratio of the re-determined target candidate geomagnetic adaptation area and each of the remaining retained candidate geomagnetic adaptation areas is determined until no candidate geomagnetic adaptation areas are retained, so as to determine the geomagnetic adaptation area corresponding to the target area.

8. A geomagnetic adaptation area positioning device for geomagnetic navigation, characterized in that: include: An acquisition module is used to acquire a geomagnetic information map corresponding to a target area, wherein the geomagnetic information map is used to characterize the geomagnetic distribution in the target area. For each grid point preset in the target area, the geomagnetic information map records geomagnetic data corresponding to the grid point, and the geomagnetic data corresponding to the grid point at least includes three-dimensional geomagnetic vector data corresponding to the grid point or combined data composed of the total geomagnetic field strength, magnetic declination and magnetic inclination corresponding to the grid point; A feature determination module, used for determining a geomagnetic feature map corresponding to the target area according to the geomagnetic information map, and inputting the geomagnetic feature map into a preset scale division model, so that the scale division model generates each scale feature map corresponding to the geomagnetic information map according to the geomagnetic feature map, and for each scale feature map, the scale feature map is used to represent the geomagnetic feature map of the geomagnetic information map at the image resolution corresponding to the scale feature map; A first determination module is used to determine, for each scale feature map, a candidate geomagnetic adaptation area in the target area at an image resolution corresponding to the scale feature map; The second determination module is used to determine the geomagnetic adaptation area corresponding to the target area according to the selected geomagnetic adaptation areas corresponding to the scale feature maps.

9. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the method described in any one of claims 1 to 7 is implemented.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the method described in any one of claims 1 to 7 is implemented.

Images