CN117705059B - Positioning method and system for remote sensing mapping image of natural resource - Google Patents

Abstract

A method and system for locating the remote-sensing mapping image of natural resource is disclosed. Firstly, remote sensing image data are acquired, then, feature description based on the region of interest is carried out on the remote sensing image data to obtain a plurality of region of interest description feature vectors, then, image positioning requirement description is acquired, then, semantic coding is carried out on the image positioning requirement description to obtain an image positioning requirement description feature vector, and finally, a positioning result is determined based on the plurality of region of interest description feature vectors and the image positioning requirement description feature vector. Therefore, the image area matched with the massive remote sensing image data can be retrieved according to semantic information contained in the image positioning requirement description input by the user, so that intelligent positioning is realized.

Description

A natural resource remote sensing mapping image positioning method and system

Technical Field

The present application relates to the field of image positioning, and more specifically, to a natural resource remote sensing mapping image positioning method and system.

Background Art

Remote sensing mapping is a technology that uses satellites, aircraft, drones and other equipment to observe and measure natural resources on the earth's surface from a distance. Remote sensing mapping can provide large-scale, high-resolution, multi-temporal image data, providing important information support for the investigation, evaluation, monitoring and management of natural resources.

However, the scale and complexity of remote sensing image data also bring challenges to image retrieval and positioning. Traditional image positioning methods have problems such as low positioning accuracy and low processing efficiency when processing large-scale and complex natural resource image data. Therefore, an optimized natural resource remote sensing mapping image positioning method and system is expected.

Summary of the invention

In order to solve the above technical problems, this application is proposed. The embodiments of this application provide a natural resource remote sensing mapping image positioning method and system, which can be combined with natural language processing technology to retrieve matching image areas from massive remote sensing image data according to the semantic information contained in the image positioning requirement description input by the user, so as to realize intelligent positioning.

According to one aspect of the present application, a natural resource remote sensing mapping image positioning method is provided, which includes:

Acquire remote sensing image data;

Performing feature description based on the region of interest on the remote sensing image data to obtain a plurality of region of interest description feature vectors;

Obtain image positioning requirement description;

semantically encoding the image positioning requirement description to obtain an image positioning requirement description feature vector; and

The positioning result is determined based on the multiple region of interest description feature vectors and the image positioning requirement description feature vector.

According to another aspect of the present application, a natural resource remote sensing mapping image positioning system is provided, comprising:

A remote sensing image data acquisition module is used to acquire remote sensing image data;

An interest feature description module is used to perform feature description based on the region of interest on the remote sensing image data to obtain a plurality of interest region description feature vectors;

An image positioning requirement description acquisition module is used to acquire an image positioning requirement description;

A semantic encoding module, used for semantically encoding the image positioning requirement description to obtain an image positioning requirement description feature vector; and

The positioning result analysis module is used to determine the positioning result based on the multiple region of interest description feature vectors and the image positioning requirement description feature vector.

Compared with the prior art, the natural resource remote sensing mapping image positioning method and system provided by the present application first obtains remote sensing image data, then performs feature description based on the region of interest on the remote sensing image data to obtain multiple region of interest description feature vectors, then obtains image positioning requirement description, then performs semantic encoding on the image positioning requirement description to obtain image positioning requirement description feature vectors, and finally determines the positioning result based on the multiple region of interest description feature vectors and the image positioning requirement description feature vectors. In this way, the image area matching the image positioning requirement description input by the user can be retrieved from the massive remote sensing image data to achieve intelligent positioning.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. The following drawings are not intentionally scaled to the actual size, and the focus is on illustrating the main purpose of the present application.

FIG. 1 is a flow chart of a natural resource remote sensing mapping image positioning method according to an embodiment of the present application.

FIG. 2 is a schematic diagram of the architecture of a natural resource remote sensing mapping image positioning method according to an embodiment of the present application.

FIG. 3 is a flowchart of sub-step S120 of the natural resource remote sensing mapping image positioning method according to an embodiment of the present application.

FIG. 4 is a flowchart of sub-step S150 of the natural resource remote sensing mapping image positioning method according to an embodiment of the present application.

FIG. 5 is a flowchart of sub-step S152 of the natural resource remote sensing mapping image positioning method according to an embodiment of the present application.

FIG. 6 is a block diagram of a natural resource remote sensing mapping image positioning system according to an embodiment of the present application.

FIG. 7 is a diagram showing an application scenario of a natural resource remote sensing mapping image positioning method according to an embodiment of the present application.

DETAILED DESCRIPTION

The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work also fall within the scope of protection of the present application.

As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and/or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and/or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. At the same time, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

In response to the above technical problems, the technical concept of the present application is to combine natural language processing technology to retrieve matching image areas from massive remote sensing image data based on the semantic information contained in the image positioning requirement description input by the user, so as to achieve intelligent positioning.

Based on this, Figure 1 is a flow chart of a natural resource remote sensing sensing and mapping image positioning method according to an embodiment of the present application. Figure 2 is a schematic diagram of the architecture of a natural resource remote sensing and mapping image positioning method according to an embodiment of the present application. As shown in Figures 1 and 2, a natural resource remote sensing and mapping image positioning method according to an embodiment of the present application includes the following steps: S110, acquiring remote sensing image data; S120, performing feature description based on the region of interest on the remote sensing image data to obtain multiple region of interest description feature vectors; S130, acquiring an image positioning requirement description; S140, semantically encoding the image positioning requirement description to obtain an image positioning requirement description feature vector; and, S150, determining a positioning result based on the multiple region of interest description feature vectors and the image positioning requirement description feature vector.

It should be understood that in step S110, remote sensing image data for positioning is obtained. Remote sensing image data is usually obtained through satellites, aircraft or other remote sensing platforms, and can provide image information of the earth's surface. In step S120, the region of interest in the remote sensing image data is characterized. The region of interest can be a region with specific attributes or targets, such as buildings, roads, water bodies, etc. By extracting the features of these regions of interest, a description feature vector of each region can be obtained. In step S130, the requirement description of image positioning is obtained, which can be a user's description of a certain target or area in the remote sensing image, such as "find the city center" or "locate the confluence of rivers". The image positioning requirement description can be a description in natural language form or other forms. In step S140, the image positioning requirement description is semantically encoded and converted into an image positioning requirement description feature vector. The semantic encoding can use natural language processing or other technologies to convert the natural language description into a computer-processable vector representation. In step S150, multiple feature vectors describing regions of interest and feature vectors describing image positioning requirements are used to determine positioning results, which can be achieved by calculating the similarity or degree of matching between the feature vectors describing regions of interest and the feature vectors describing image positioning requirements. Based on the similarity or degree of matching, it can be determined which regions of interest have a higher degree of matching with the image positioning requirements, thereby obtaining positioning results. The combination of these steps can help achieve remote sensing image positioning. By extracting, encoding and matching features of remote sensing image data with image positioning requirements, the location of the region of interest and the positioning results can be determined.

Specifically, in the technical solution of the present application, remote sensing image data is first acquired; and multiple regions of interest are extracted from the remote sensing image data. The regions of interest are regions with specific features in the remote sensing image data, and these regions of interest may include specific types of land objects, landforms, vegetation, water bodies, etc. By extracting multiple regions of interest from the remote sensing image data, the retrieval scope of the subsequent model can be narrowed, and a large amount of interference information and noise can be avoided to have an adverse effect on the positioning results.

In a specific example of the present application, multiple regions of interest can be extracted from the remote sensing image data by manual capture. Specifically, a professional remote sensing image interpreter can use image processing software to manually select regions of interest in the remote sensing image. This method requires manual participation and has certain requirements on the operator's professional knowledge and experience, but it can accurately select regions with specific features. In another specific example of the present application, multiple regions of interest can be extracted from the remote sensing image data using computer vision and image processing technology. For example, by constructing a target recognition network and other methods.

Then, the multiple regions of interest are respectively passed through a feature descriptor based on a convolutional neural network model to obtain multiple region of interest description feature vectors. That is, the high-level semantic features of each region of interest are extracted using the multi-layer convolution and pooling operations of the convolutional neural network model. Among them, these high-level semantic features can better describe the morphology, texture, structure and other features of the region.

Accordingly, in step S120, as shown in FIG3 , the remote sensing image data is subjected to feature description based on the region of interest to obtain a plurality of region of interest description feature vectors, including: S121, extracting a plurality of regions of interest from the remote sensing image data; and, S122, performing feature extraction and feature description on the plurality of regions of interest using a deep learning network model to obtain the plurality of region of interest description feature vectors.

Wherein, in step S122, the deep learning network model is a feature descriptor based on a convolutional neural network model; wherein the feature descriptor based on a convolutional neural network model includes an input layer, a convolution layer, an activation layer, a pooling layer, and an output layer. Specifically, using the deep learning network model to perform feature extraction and feature description on the multiple regions of interest to obtain the multiple regions of interest description feature vectors includes: passing the multiple regions of interest through the feature descriptor based on the convolutional neural network model respectively to obtain the multiple regions of interest description feature vectors.

It should be understood that Convolutional Neural Network (CNN) is a deep learning network model that is mainly used to process data with a grid structure. It is widely used in the field of computer vision and can effectively capture spatial local features in images with translation invariance. The main components of the convolutional neural network model are as follows: 1. Input layer: accepts images or regions of interest as input. 2. Convolution layer: extracts features from the input data through a series of convolution operations. The convolution operation uses a convolution kernel (also called a filter) to perform a sliding window calculation on the input data to capture local features. 3. Activation layer: introduces nonlinear transformations to increase the expressive power of the network. Common activation functions include the ReLU (Rectified Linear Unit) function. 4. Pooling layer: reduces the size of the feature map through downsampling operations while retaining important features. Common pooling operations include maximum pooling and average pooling. 5. Output layer: converts the feature map obtained through convolution, activation and pooling operations into the final feature vector representation. The advantage of the convolutional neural network model is that it can automatically learn the feature representation in the image without manually designing the feature extractor. Through the training process, the network can learn features that are useful for the task, thereby achieving feature extraction and description of the region of interest. In remote sensing image positioning, the feature descriptor based on the convolutional neural network model can extract the semantic features of the region of interest for subsequent positioning result determination.

At the same time, the image positioning requirement description is obtained; and the image positioning requirement description is semantically encoded to obtain an image positioning requirement description feature vector. Here, the image positioning requirement description is a natural language description, which is usually the way humans understand and express information, but computers cannot directly understand and recognize text data. By semantically encoding the image positioning requirement description, natural language can be converted into a numerical feature vector that can be processed by a computer, and the semantic information in the image positioning requirement description can be extracted to accurately understand the specific requirements and expectations of the image positioning task, and clarify the positioning objectives and scope.

Next, the projection layer is used to fuse the image positioning requirement description feature vector and each of the multiple region of interest description feature vectors to obtain multiple semantic matching feature vectors. That is, the image positioning requirement description feature vector and each region of interest description feature vector are mapped to the same semantic feature space through the projection layer for semantic matching.

Furthermore, the multiple semantic matching feature vectors are passed through a classifier to obtain multiple classification results, and each classification result is used to indicate whether the matching degree exceeds a predetermined threshold. In the actual application scenario of the present application, each region of interest corresponding to each corrected semantic matching feature vector whose matching degree exceeds a predetermined threshold is used as a positioning region to obtain multiple positioning regions; and the multiple positioning regions are used as the positioning results.

Accordingly, in step S150, as shown in FIG4 , the positioning result is determined based on the multiple region of interest description feature vectors and the image positioning requirement description feature vector, including: S151, fusing the image positioning requirement description feature vector and each region of interest description feature vector in the multiple region of interest description feature vectors to obtain multiple semantic matching feature vectors; and, S152, determining the positioning result that meets the image positioning requirement description based on the multiple semantic matching feature vectors.

It should be understood that in step S151, the feature vector describing the image positioning requirement is fused with the feature vector describing each region of interest to obtain multiple semantic matching feature vectors. The fusion can be performed in different ways, such as concatenating two feature vectors, weighted summing, or using other fusion strategies to comprehensively consider the image positioning requirement and the characteristics of the region of interest. In step S152, multiple semantic matching feature vectors are used to determine the final positioning result to meet the image positioning requirement description. A similarity metric or other matching algorithm can be used to evaluate the degree of matching between each semantic matching feature vector and the image positioning requirement. According to the degree of matching, the region of interest corresponding to the feature vector with the highest matching degree can be selected as the positioning result, or further decision-making and reasoning can be performed to determine the final positioning result. Through the combination of these two steps, the image positioning requirement can be fused and matched with the characteristics of the region of interest, so as to obtain a positioning result that meets the requirements. This fusion and matching process can help determine the degree of semantic matching between the region of interest and the image positioning requirement, thereby providing an accurate positioning result.

Among them, in step S151, the image positioning requirement description feature vector and each region of interest description feature vector in the multiple region of interest description feature vectors are fused to obtain multiple semantic matching feature vectors, including: using a projection layer to respectively fuse the image positioning requirement description feature vector and each region of interest description feature vector in the multiple region of interest description feature vectors to obtain the multiple semantic matching feature vectors.

It is worth mentioning that the projection layer is a layer in the neural network that is used to map input data from one feature space to another feature space. In this case, the projection layer is used to map the image positioning requirement description feature vector and the region of interest description feature vector to obtain a semantic matching feature vector. Specifically, the projection layer can be a fully connected layer, also known as a linear layer or a dense layer. Each neuron in the fully connected layer is connected to all neurons in the previous layer and linearly transformed through weights and biases. In this case, the projection layer inputs the image positioning requirement description feature vector and the region of interest description feature vector into two fully connected layers respectively, and maps them to another feature space through the learned weights and biases. Through the mapping of the projection layer, features in different feature spaces can be fused to obtain multiple semantic matching feature vectors. These feature vectors will contain semantic information about the image positioning requirements and the region of interest, which will help determine the subsequent matching and positioning results. The design and parameter learning of the projection layer can be adjusted and optimized according to the specific tasks and data to obtain the best fusion effect.

In a specific example of the present application, a projection layer is used to fuse the image positioning requirement description feature vector and each of the plurality of region of interest description feature vectors to obtain the plurality of semantic matching feature vectors, including: using the following projection formula to fuse the image positioning requirement description feature vector and each of the plurality of region of interest description feature vectors to obtain the plurality of semantic matching feature vectors; wherein the projection formula is:

Wherein, _Vf is the semantic matching feature vector, _V1 is the image positioning requirement description feature vector, _V2 is the feature vector describing each region of interest, [·;·] represents cascade, Represents a projection map of a vector.

Here, the feature vector describing the image positioning requirements and the feature vector describing the region of interest are mapped into the same semantic feature space through a shared projection layer, so as to constrain the high-dimensional feature distribution manifolds of the two to the same measurement scale, so that they can be directly compared and matched.

Further, in step S152, as shown in Figure 5, the positioning result that meets the image positioning requirement description is determined based on the multiple semantic matching feature vectors, including: S1521, performing feature distribution correction on the multiple semantic matching feature vectors to obtain multiple corrected semantic matching feature vectors; S1522, passing the multiple corrected semantic matching feature vectors through a classifier to obtain multiple classification results, and each classification result is used to indicate whether the matching degree exceeds a predetermined threshold; S1523, taking each of the regions of interest corresponding to each of the corrected semantic matching feature vectors whose matching degree exceeds the predetermined threshold as a positioning area to obtain multiple positioning areas; and, S1524, taking the multiple positioning areas as the positioning results.

It should be understood that in step S1521, feature distribution correction is performed on multiple semantic matching feature vectors. The purpose of feature distribution correction is to make their distribution in the feature space more consistent or more consistent with a certain expected distribution by normalizing, standardizing or other processing of the feature vectors, so as to improve the accuracy of comparison and matching between feature vectors. In step S1522, the corrected semantic matching feature vector is input into the classifier for classification. The classifier can be a binary classifier for judging whether the matching degree of the feature vector exceeds a predetermined threshold. The classifier can be trained based on a machine learning algorithm to judge whether the feature vector meets the matching conditions by learning the labeled sample data. The classification result can be binary (match/mismatch) or a probability value, indicating the confidence of the match. In step S1523, according to the corrected semantic matching feature vector whose matching degree exceeds the predetermined threshold, the corresponding region of interest is determined as the positioning region. These positioning regions represent the regions that match the image positioning requirement description and can be considered as possible positioning results. In step S1524, multiple positioning areas are used as the final positioning results. These positioning areas are screened and determined according to the matching degree and threshold in the previous steps, representing possible locations that match the image positioning requirement description. Using these positioning areas as positioning results can provide multiple candidate locations for further analysis and decision-making. Through the combination of these steps, the positioning results that meet the image positioning requirement description can be determined based on the correction, classification and threshold judgment of the semantic matching feature vector. This method can improve the accuracy and robustness of positioning, while providing multiple possible positioning areas for selection and further processing.

In the above technical solution, the image positioning requirement description feature vector is used to express the encoded text semantic features of the image positioning requirement description, and the multiple regions of interest description feature vectors respectively express the image semantic features of the multiple regions of interest. Therefore, when using the projection layer to fuse the image positioning requirement description feature vector and each region of interest description feature vector in the multiple regions of interest description feature vectors respectively, it is considered that the cross-modal semantic feature difference between the image positioning requirement description feature vector and the region of interest description feature vector may lead to the mapping sparsity of semantic features under the shared dimension, thereby affecting the expression effect of the obtained multiple semantic matching feature vectors. Therefore, it is expected to optimize the feature projection mapping based on the feature expression significance and criticality of each of the image positioning requirement description feature vector and the region of interest description feature vector, so as to improve the expression effect of the multiple semantic matching feature vectors. Based on this, the applicant of this application corrects the image positioning requirement description feature vector and each region of interest description feature vector.

Accordingly, in step S1521, feature distribution correction is performed on the multiple semantic matching feature vectors to obtain multiple corrected semantic matching feature vectors, including: calculating multiple corrected feature vectors of the image positioning requirement description feature vector and each of the multiple region of interest description feature vectors using the following correction formula; wherein the correction formula is:

Wherein, _V1 is the image positioning requirement description feature vector, and _V2 is each region of interest description feature vector in the plurality of region of interest description feature vectors, represents the position-by-position square root of the eigenvector, v _1max ^-1 and v _2max ^-1 are the inverses of the maximum eigenvalues of the eigenvectors V ₁ and V ₂ , respectively, α and β are weight hyperparameters, ⊙ represents the position-by-position point multiplication, represents vector subtraction, V _c is each corrected feature vector in the multiple corrected feature vectors; and the multiple semantic matching feature vectors and the multiple corrected feature vectors are respectively fused to obtain the multiple corrected semantic matching feature vectors.

Here, the pre-segmented local group of the feature value set is obtained by taking the square root of each feature value of the feature vector describing the image positioning requirements and the feature vector describing the region of interest, and regressing the key maximum features of the feature vector describing the image positioning requirements and the feature vector describing the region of interest from them. In this way, the location-based significant distribution of the feature value can be improved based on the idea of farthest point sampling, so that the sparse correspondence between feature vectors can be controlled through key features with significant distribution, so as to achieve the restoration of the original manifold geometry of the feature vector describing the image positioning requirements and the feature vector describing the region of interest by the correction feature vector V _c . In this way, the correction feature vector V _c is fused with the semantic matching feature vector, so as to improve the expression effect of the semantic matching feature vector, thereby improving the accuracy of the classification result obtained by the classifier.

Further, in step S1522, the multiple corrected semantic matching feature vectors are passed through a classifier to obtain multiple classification results, and each classification result is used to indicate whether the matching degree exceeds a predetermined threshold, including: using the fully connected layer of the classifier to fully connect encode the multiple corrected semantic matching feature vectors respectively to obtain multiple encoded classification feature vectors; and, inputting the multiple encoded classification feature vectors into the Softmax classification function of the classifier respectively to obtain the multiple classification results.

It should be understood that the role of the classifier is to use the given categories and known training data to learn classification rules and classifiers, and then classify (or predict) unknown data. Logistic regression and SVM are often used to solve binary classification problems. For multi-class classification problems, logistic regression or SVM can also be used, but multiple binary classifications are required to form a multi-classification problem. However, this is prone to errors and is not efficient. Commonly used multi-classification methods include the Softmax classification function.

It is worth mentioning that fully connected encoding refers to the process of encoding input data through a fully connected layer. A fully connected layer is a common layer type in a neural network, and each of its neurons is connected to all neurons in the previous layer. In fully connected encoding, each feature of the input data is connected to each neuron in the fully connected layer, and the output of the neuron is calculated through a combination of weights and biases. The role of fully connected encoding is to convert the input data into a higher-level representation to extract richer features in the data. Through the nonlinear transformation and parameter learning of the fully connected layer, the complex relationships and patterns in the input data can be captured. By inputting the corrected semantic matching feature vector into the fully connected layer for encoding, the feature vector can be converted into a more expressive encoded classification feature vector. After these encoded classification feature vectors pass through the Softmax classification function, multiple classification results can be obtained to indicate whether the matching degree exceeds the predetermined threshold. Through the combination of fully connected encoding and classifiers, semantic matching can be classified and judged more accurately.

In summary, the natural resource remote sensing mapping image positioning method based on the embodiment of the present application is explained, which can retrieve matching image areas from massive remote sensing image data according to the semantic information contained in the image positioning requirement description input by the user to achieve intelligent positioning.

FIG6 is a block diagram of a natural resource remote sensing image positioning system 100 according to an embodiment of the present application. As shown in FIG6, the natural resource remote sensing image positioning system 100 according to an embodiment of the present application includes: a remote sensing image data acquisition module 110 for acquiring remote sensing image data; an interest feature description module 120 for performing feature description based on the region of interest on the remote sensing image data to obtain multiple interest region description feature vectors; an image positioning requirement description acquisition module 130 for acquiring image positioning requirement description; a semantic encoding module 140 for performing semantic encoding on the image positioning requirement description to obtain an image positioning requirement description feature vector; and a positioning result analysis module 150 for determining the positioning result based on the multiple interest region description feature vectors and the image positioning requirement description feature vector.

In one example, in the above-mentioned natural resource remote sensing mapping image positioning system 100, the feature description module 120 of interest includes: a region of interest extraction unit, used to extract multiple regions of interest from the remote sensing image data; and a feature extraction and description unit, used to use a deep learning network model to perform feature extraction and feature description on the multiple regions of interest to obtain the multiple region of interest description feature vectors.

Here, those skilled in the art can understand that the specific functions and operations of each module in the above-mentioned natural resource remote sensing mapping image positioning system 100 have been described in detail in the description of the natural resource remote sensing mapping image positioning method with reference to Figures 1 to 5 above, and therefore, its repeated description will be omitted.

As described above, the natural resource remote sensing image positioning system 100 according to the embodiment of the present application can be implemented in various wireless terminals, such as a server with a natural resource remote sensing image positioning algorithm. In one example, the natural resource remote sensing image positioning system 100 according to the embodiment of the present application can be integrated into the wireless terminal as a software module and/or a hardware module. For example, the natural resource remote sensing image positioning system 100 can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the natural resource remote sensing image positioning system 100 can also be one of the many hardware modules of the wireless terminal.

Alternatively, in another example, the natural resource remote sensing mapping image positioning system 100 and the wireless terminal may also be separate devices, and the natural resource remote sensing mapping image positioning system 100 may be connected to the wireless terminal via a wired and/or wireless network and transmit interactive information in accordance with an agreed data format.

Figure 7 is an application scenario diagram of the natural resource remote sensing image positioning method according to an embodiment of the present application. As shown in Figure 7, in this application scenario, first, remote sensing image data (e.g., D1 shown in Figure 7) and image positioning requirement description (e.g., D2 shown in Figure 7) are obtained, and then the remote sensing image data and the image positioning requirement description are input into a server (e.g., S shown in Figure 7) deployed with a natural resource remote sensing image positioning algorithm, wherein the server can use the natural resource remote sensing image positioning algorithm to process the remote sensing image data and the image positioning requirement description to obtain multiple classification results for indicating whether the matching degree exceeds a predetermined threshold.

According to another aspect of the present application, a non-volatile computer-readable storage medium is provided, on which computer-readable instructions are stored. When the instructions are executed by a computer, the above-mentioned method can be executed.

The program part in the technology can be considered as a "product" or "manufactured product" in the form of executable code and/or related data, which is participated in or realized by computer-readable media. Tangible and permanent storage media can include any memory or storage used by any computer, processor, or similar device or related module. For example, various semiconductor memories, tape drives, disk drives or any similar devices that can provide storage functions for software.

All software or parts thereof may sometimes be communicated over a network, such as the Internet or other communications network. Such communications can load software from one computer device or processor to another. As used herein, unless restricted to tangible "storage" media, other terms referring to computer or machine "readable media" refer to media that participate in the process of executing any instructions by the processor.

In addition, it will be appreciated by those skilled in the art that various aspects of the present application may be illustrated and described by a number of patentable categories or situations, including any new and useful process, machine, product or combination of substances, or any new and useful improvements thereto. Accordingly, various aspects of the present application may be performed entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. The above hardware or software may all be referred to as "data blocks", "modules", "engines", "units", "components" or "systems". In addition, various aspects of the present application may be represented as a computer product located in one or more computer-readable media, which includes computer-readable program code.

Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those of ordinary skill in the art to which this application belongs. It should also be understood that terms such as those defined in common dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an idealized or extremely formal sense, unless explicitly defined as such herein.

The above is an explanation of the present application and should not be considered as a limitation thereof. Although several exemplary embodiments of the present application are described, it will be readily understood by those skilled in the art that many modifications may be made to the exemplary embodiments without departing from the novel teachings and advantages of the present application. Therefore, all of these modifications are intended to be included within the scope of the present application as defined by the claims. It should be understood that the above is an explanation of the present application and should not be considered to be limited to the specific embodiments disclosed, and modifications to the disclosed embodiments and other embodiments are intended to be included within the scope of the appended claims. The present application is defined by the claims and their equivalents.

Claims (8)

1. A natural resource remote sensing mapping image positioning method, characterized by comprising: Acquire remote sensing image data; Performing feature description based on the region of interest on the remote sensing image data to obtain a plurality of region of interest description feature vectors; Obtain image positioning requirement description; semantically encoding the image positioning requirement description to obtain an image positioning requirement description feature vector; and Determine a positioning result based on the multiple regions of interest description feature vectors and the image positioning requirement description feature vector; Determining a positioning result based on the multiple regions of interest description feature vectors and the image positioning requirement description feature vector includes: Fusion of the image positioning requirement description feature vector and each of the plurality of region of interest description feature vectors to obtain a plurality of semantic matching feature vectors; Determine the positioning result that meets the image positioning requirement description based on the multiple semantic matching feature vectors; Determining the positioning result that meets the image positioning requirement description based on the multiple semantic matching feature vectors includes: Performing feature distribution correction on the plurality of semantic matching feature vectors to obtain a plurality of corrected semantic matching feature vectors; Passing the multiple corrected semantic matching feature vectors through a classifier to obtain multiple classification results, each of which is used to indicate whether the matching degree exceeds a predetermined threshold; Taking each of the regions of interest corresponding to each of the corrected semantic matching feature vectors whose matching degree exceeds a predetermined threshold as a positioning region to obtain a plurality of positioning regions; Taking the multiple positioning areas as the positioning results; Performing feature distribution correction on the multiple semantic matching feature vectors to obtain multiple corrected semantic matching feature vectors includes: calculating multiple corrected feature vectors of the image positioning requirement description feature vector and each of the multiple region of interest description feature vectors using the following correction formula; wherein the correction formula is: Wherein, _V1 is the image positioning requirement description feature vector, and _V2 is each region of interest description feature vector in the plurality of region of interest description feature vectors, represents the position-by-position square root of the eigenvector, v _1max ^-1 and v _2max ^-1 are the inverses of the maximum eigenvalues of the eigenvectors V ₁ and V ₂ , respectively, α and β are weight hyperparameters, ⊙ represents the position-by-position point multiplication, represents vector subtraction, V _c is each corrected feature vector in the multiple corrected feature vectors; and the multiple semantic matching feature vectors and the multiple corrected feature vectors are respectively fused to obtain the multiple corrected semantic matching feature vectors. 2. The natural resource remote sensing mapping image positioning method according to claim 1, characterized in that the remote sensing image data is subjected to feature description based on the region of interest to obtain a plurality of region of interest description feature vectors, comprising: Extracting a plurality of regions of interest from the remote sensing image data; and A deep learning network model is used to perform feature extraction and feature description on the multiple regions of interest to obtain the multiple region of interest description feature vectors. 3. The natural resource remote sensing mapping image positioning method according to claim 2, characterized in that the deep learning network model is a feature descriptor based on a convolutional neural network model; Among them, the feature descriptor based on the convolutional neural network model includes an input layer, a convolution layer, an activation layer, a pooling layer and an output layer. 4. The natural resource remote sensing mapping image positioning method according to claim 3 is characterized in that the deep learning network model is used to extract and describe the features of the multiple regions of interest to obtain the multiple region of interest description feature vectors, including: The multiple regions of interest are respectively passed through the feature descriptor based on the convolutional neural network model to obtain the multiple region of interest description feature vectors. 5. The natural resource remote sensing mapping image positioning method according to claim 4, characterized in that the image positioning requirement description feature vector and each of the plurality of region of interest description feature vectors are fused to obtain a plurality of semantic matching feature vectors, comprising: The projection layer is used to fuse the image positioning requirement description feature vector and each of the plurality of interest region description feature vectors to obtain the plurality of semantic matching feature vectors. 6. The natural resource remote sensing mapping image positioning method according to claim 5, characterized in that the projection layer is used to fuse the image positioning requirement description feature vector and each of the plurality of interest region description feature vectors to obtain the plurality of semantic matching feature vectors, including: The image positioning requirement description feature vector and each of the plurality of region of interest description feature vectors are respectively fused with the following projection formula to obtain the plurality of semantic matching feature vectors; Wherein, the projection formula is: Wherein, _Vf is the semantic matching feature vector, _V1 is the image positioning requirement description feature vector, _V2 is the feature vector describing each region of interest, [·;·] represents cascade, Represents a projection map of a vector. 7. A natural resource remote sensing mapping image positioning system, used to execute the natural resource remote sensing mapping image positioning method according to claim 1, characterized in that it comprises: A remote sensing image data acquisition module is used to acquire remote sensing image data; An interest feature description module is used to perform feature description based on the region of interest on the remote sensing image data to obtain a plurality of interest region description feature vectors; An image positioning requirement description acquisition module is used to acquire an image positioning requirement description; A semantic encoding module, used for semantically encoding the image positioning requirement description to obtain an image positioning requirement description feature vector; and The positioning result analysis module is used to determine the positioning result based on the multiple region of interest description feature vectors and the image positioning requirement description feature vector. 8. The natural resource remote sensing mapping image positioning system according to claim 7, characterized in that the interesting feature description module comprises: A region of interest extraction unit, used to extract multiple regions of interest from the remote sensing image data; and A feature extraction and description unit is used to use a deep learning network model to perform feature extraction and feature description on the multiple regions of interest to obtain the multiple region of interest description feature vectors.