Method, device, equipment, and medium for automatic interpretation of remote sensing images based on deep learning
By preprocessing and fusion classification of remote sensing images based on deep learning, the problems of low efficiency and low accuracy of traditional remote sensing images are solved, and more efficient and accurate interpretation of remote sensing images are achieved.
Patent Information
- Application Number
- CN202411079150.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-07
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2044-08-07
AI Technical Summary
Traditional remote sensing image interpretation methods are inefficient and have great subjectivity, resulting in inconsistent interpretation results and high error rate.
Using a deep learning-based method, the multi-stage remote sensing images of the same sensor satellite in different regions and periods are preprocessed, and the pre-trained convolutional neural network is called for multi-layer convolution and pooling operations to generate feature maps, and the The feature map is fused and classified, and classification results are generated and integrated output is performed, including assigning colors and labeling boundaries to the perturbing map.
It improves the efficiency and accuracy of remote sensing image interpretation, reduces the error of manual interpretation, and achieves more consistent and reliable interpretation results.
Smart Images

Figure CN118608988B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of remote sensing image interpretation, and in particular to a method, device, equipment, and medium for automatic remote sensing image interpretation based on deep learning. Background Art
[0002] Remote sensing images have important applications in environmental monitoring, resource exploration, urban planning, etc. In environmental monitoring, remote sensing images can be used to monitor changes in forest cover, wetland protection, land use changes, water resource management, etc. In resource exploration, remote sensing images can help identify the distribution of mineral resources, evaluate agricultural resources, monitor fishery resources, etc. In urban planning, remote sensing images can provide key data support for urban expansion, transportation networks, infrastructure construction, etc.
[0003] Traditional remote sensing image interpretation methods mainly rely on manual analysis, that is, professionals visually inspect and manually mark features in the image. This method is not only inefficient, but also easily affected by human factors and has a large degree of subjectivity. For example, different analysts may produce different results for the same image, resulting in inconsistent interpretation results. In addition, manual interpretation is also easily affected by factors such as fatigue and lack of experience, further increasing the error rate in the interpretation process.
[0004] In view of this, this application is filed. Summary of the invention
[0005] The present invention discloses a method, device, equipment, and medium for automatic interpretation of remote sensing images based on deep learning, aiming to improve the efficiency and accuracy of remote sensing image interpretation.
[0006] The first embodiment of the present invention provides a method for automatic interpretation of remote sensing images based on deep learning, which is characterized by comprising:
[0007] Acquire multiple remote sensing images of the same sensor satellite in different regions and at different times, and pre-process the remote sensing images;
[0008] Calling a pre-trained convolutional neural network to perform multi-layer convolution and pooling operations on the multi-period remote sensing images to generate feature maps at different levels, and fusing the feature maps at different levels into a comprehensive representation of the ground object target, wherein the feature map includes the edge, texture, shape and structure of the remote sensing image;
[0009] A pre-trained classifier is called to classify the comprehensive representation of the ground object target to generate a classification result, and the classification result is integrated and output in a table form, wherein the integration operation includes assigning different colors to each type of disturbance pattern of the classification result, marking the disturbance pattern with different shapes, and depicting the boundary of each disturbance pattern.
[0010] Preferably, the preprocessing includes radiation calibration, geometric correction, and atmospheric correction.
[0011] Preferably, the expression for generating feature maps at different levels is:
[0012]
[0013] in, It is a feature map In Location and The value on the convolution kernel, It is remote sensing image In Location and The pixel values on each channel are It is The convolution kernels are located at and The weights on the channels, , is the width and height of the convolution kernel, and It is the center position of the convolution kernel on the remote sensing image.
[0014] Preferably, the method further comprises: constructing a training set for training a convolutional neural network, specifically:
[0015] Obtain the information of various disturbance spots collected on site, and organize and archive the information of various disturbance spots to establish an interpretation identification library, wherein the information of various disturbance spots includes specific location and type information,
[0016] Corresponding spots are drawn one by one on the collected remote sensing images according to the information in the interpretation identification library, and corresponding attribute information is added to each spot to form a training set, wherein the attribute information includes type, location, range size and description.
[0017] The second embodiment of the present invention provides a remote sensing image automatic interpretation device based on deep learning, comprising:
[0018] A preprocessing unit, used to obtain multiple remote sensing images of the same sensor satellite in different regions and at different times, and to preprocess the remote sensing images;
[0019] A feature fusion unit, used to call a pre-trained convolutional neural network to perform multi-layer convolution and pooling operations on the multi-period remote sensing images to generate feature maps at different levels, and fuse the feature maps at different levels into a comprehensive representation of the ground object target, wherein the feature map includes the edge, texture, shape and structure of the remote sensing image;
[0020] A classification unit is used to call a pre-trained classifier to classify the comprehensive representation of the ground object target to generate a classification result, and output the classification result in a table after integration operation, wherein the integration operation includes assigning different colors to each type of disturbance pattern of the classification result, marking the disturbance pattern with different shapes, and depicting the boundary of each disturbance pattern.
[0021] Preferably, the preprocessing includes radiation calibration, geometric correction, and atmospheric correction.
[0022] Preferably, the expression for generating feature maps at different levels is:
[0023]
[0024] in, It is a feature map In Location and The value on the convolution kernel, It is remote sensing image In Location and The pixel values on each channel are It is The convolution kernels are located at and The weights on the channels, , is the width and height of the convolution kernel, and It is the center position of the convolution kernel on the remote sensing image.
[0025] Preferably, the method further comprises: constructing a training set for training a convolutional neural network, specifically:
[0026] Obtain the information of various disturbance spots collected on site, and organize and archive the information of various disturbance spots to establish an interpretation identification library, wherein the information of various disturbance spots includes specific location and type information,
[0027] Corresponding spots are drawn one by one on the collected remote sensing images according to the information in the interpretation identification library, and corresponding attribute information is added to each spot to form a training set, wherein the attribute information includes type, location, range size and description.
[0028] The third embodiment of the present invention provides an automatic interpretation device for remote sensing images based on deep learning, including a memory and a processor, wherein the memory stores a computer program, and the computer program can be executed by the processor to implement an automatic interpretation method for remote sensing images based on deep learning as described in any one of the above.
[0029] The fourth embodiment of the present invention provides a computer-readable storage medium, characterized in that a computer program is stored therein, and the computer program can be executed by a processor of a device where the computer-readable storage medium is located to implement a method for automatic interpretation of remote sensing images based on deep learning as described in any one of the above items.
[0030] A method, device, equipment, and medium for automatic interpretation of remote sensing images based on deep learning provided by the present invention first obtain multi-period remote sensing images of the same sensor satellite in different regions and different periods, and pre-process the remote sensing images; then, call a pre-trained convolutional neural network to perform multi-layer convolution and pooling operations on the multi-period remote sensing images to generate feature maps of different levels, and fuse the feature maps of different levels into a comprehensive representation of ground objects; finally, call a pre-trained classifier to classify the comprehensive representation of the ground objects to generate a classification result, and integrate the classification results and output them in the form of a table, wherein the integration operation includes assigning different colors to each type of disturbance spots of the classification results, marking the disturbance spots with different shapes, and depicting the boundaries of each disturbance spot, thereby solving the problems of low efficiency and low accuracy of manual interpretation. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] Figure 1 It is a flowchart of a method for automatic interpretation of remote sensing images based on deep learning provided by the first embodiment of the present invention;
[0032] Figure 2 It is a module schematic diagram of a remote sensing image automatic interpretation device based on deep learning provided by the second embodiment of the present invention. DETAILED DESCRIPTION
[0033] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0034] In order to better understand the technical solution of the present invention, the embodiments of the present invention are described in detail below with reference to the accompanying drawings.
[0035] It should be clear that the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0036] The terms used in the embodiments of the present invention are only for the purpose of describing specific embodiments, and are not intended to limit the present invention. The singular forms "a", "said" and "the" used in the embodiments of the present invention and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings.
[0037] It should be understood that the term "and / or" used in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.
[0038] The word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrases "if it is determined" or "if (stated condition or event) is detected" may be interpreted as "when it is determined" or "in response to determining" or "when detecting (stated condition or event)" or "in response to detecting (stated condition or event)", depending on the context.
[0039] The "first\second" mentioned in the embodiments is only to distinguish similar objects, and does not represent a specific order for the objects. It is understandable that the "first\second" can be interchanged with the specific order or sequence where permitted. It should be understood that the objects distinguished by "first\second" can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than those illustrated or described herein.
[0040] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.
[0041] The present invention discloses a method, device, equipment, and medium for automatic interpretation of remote sensing images based on deep learning, aiming to improve the efficiency and accuracy of remote sensing image interpretation.
[0042] The first embodiment of the present invention provides a method for automatic interpretation of remote sensing images based on deep learning, characterized in that it can be performed by a remote sensing image automatic interpretation device based on deep learning (hereinafter referred to as the interpretation device), and in particular, is performed by one or more processors in the interpretation device to implement at least the following steps:
[0043] S101, acquiring multiple remote sensing images of the same sensor satellite in different regions and at different times, and preprocessing the remote sensing images;
[0044] In this embodiment, the interpretation device can be a terminal with data processing capabilities such as a desktop computer, a laptop computer, a workstation, a server, etc. The interpretation device can be installed with a corresponding operating system and application software, and the functions required by this embodiment can be implemented through the combination of the operating system and application software.
[0045] In this embodiment, the remote sensing images can be obtained by obtaining satellite images from major remote sensing image data platforms, and the remote sensing images of the required time period and area can be downloaded in batches through an online platform or a data interface, and the same sensor on the same satellite is used to obtain data to ensure the uniformity and comparability of the data;
[0046] In a possible embodiment of the present invention, DEM data is obtained while acquiring remote sensing images, wherein the DEM data provides surface height information for reflecting the undulations of the terrain. In the process of geometric correction of the remote sensing images, the DEM data can be used to correct the geometric distortion in the images to ensure that the images are consistent with the actual terrain position. Specifically, in this embodiment, the preprocessing includes radiation calibration, geometric correction, and atmospheric correction.
[0047] Radiometric calibration is to convert raw image data into physical quantities, such as radiance or reflectivity, to eliminate the effects of differences between sensors and time changes. It can use the sensor's own radiometric calibration parameters and combine them with ground observation data to perform radiometric correction on the image. Common methods include absolute radiometric calibration and relative radiometric calibration.
[0048] Geometric correction is to correct the geometric distortion in the image so that the spatial position of the image is consistent with the geographic coordinates, thereby ensuring accurate spatial comparison and superposition of images from different times and regions. It can use geographic reference points (GCP) and digital elevation models (DEM) for geometric correction to eliminate the geometric deformation of the image caused by sensor tilt, terrain undulations and earth curvature.
[0049] Atmospheric correction is to remove the influence of the atmosphere on remote sensing images, such as the radiation scattering and absorption effects caused by aerosols, clouds, water vapor, etc., so as to improve the authenticity and contrast of the images. Atmospheric radiation transmission models, such as the 6S (Second Simulation of a Satellite Signal in the Solar Spectrum) model, can be used to combine the metadata of the image and ground observation data to perform atmospheric correction on the image. The specific steps include atmospheric absorption correction, atmospheric scattering correction and surface reflectivity calculation.
[0050] S102, calling a pre-trained convolutional neural network to perform multi-layer convolution and pooling operations on the multi-period remote sensing images to generate feature maps at different levels, and fusing the feature maps at different levels into a comprehensive representation of the ground object target, wherein the feature map includes the edge, texture, shape and structure of the remote sensing image;
[0051] The expression for generating feature maps at different levels is:
[0052]
[0053] in, It is a feature map In Location and The value on the convolution kernel, It is remote sensing image In Location and The pixel values on each channel are It is The convolution kernels are located at and The weights on the channels, , is the width and height of the convolution kernel, and It is the center position of the convolution kernel on the remote sensing image.
[0054] It should be noted that in each layer of the convolutional neural network, the convolution operation scans the input image through multiple convolution kernels (filters) to extract local features (multi-layer convolution operations can effectively capture feature information of different scales in the image). These convolution kernels can capture low-level features such as edges, textures, shapes, and structures in remote sensing images. The superposition of convolution layers allows feature extraction to gradually deepen, thereby obtaining higher-level and more abstract feature representations;
[0055] The pooling layer downsamples the feature map to reduce its size while retaining important feature information. Commonly used pooling operations include Max Pooling and Average Pooling. Pooling operations can effectively reduce computational complexity and prevent model overfitting.
[0056] Furthermore, in this embodiment, the feature vector after feature fusion can be input into one or more fully connected layers (FC layers for short). A fully connected layer is a neural network layer in which all input neurons are connected to all output neurons, and is used to map high-dimensional feature vectors to classification space. Through the linear transformation of the fully connected layer and the nonlinear transformation of the activation function (such as ReLU, Sigmoid, etc.), the feature vector is further processed and reduced to a feature space suitable for classification.
[0057] S103, calling a pre-trained classifier to classify the comprehensive representation of the ground object target to generate a classification result, and integrating the classification result and outputting it in the form of a table, wherein the integration operation includes assigning different colors to each type of disturbance pattern of the classification result, marking the disturbance pattern with different shapes, and depicting the boundary of each disturbance pattern.
[0058] It should be noted that the pre-trained classifier is a model that has been trained on a large amount of data and can identify and classify specific targets. In this embodiment, a support vector machine (SVM) can be used to perform the classification task. The fused feature map is input into the pre-trained classifier, and the classifier identifies and classifies the comprehensive representation of each ground object target to generate a classification result. These classification results include the identification and labeling of different ground object types, such as vegetation, water bodies, buildings, etc.
[0059] Furthermore, in this embodiment, different colors can be assigned to each type of disturbance pattern. For example, vegetation is marked as green, water bodies are marked as blue, buildings are marked as red, and so on. Color allocation can intuitively distinguish different types of ground objects, making the classification results clearer and convenient for users to quickly identify and analyze. Use different shapes to mark the disturbance patterns, for example, use circles to mark vegetation, squares to mark buildings, triangles to mark water bodies, and so on. Shape annotation further enhances the visualization of the classification results, making different types of objects more obvious in the image, which is convenient for further analysis and application. In addition, the boundaries of each disturbance pattern are depicted to make the range and outline of the ground object clearer. Boundary depiction can accurately display the shape and position of the ground object, improve the accuracy of the interpretation results, and facilitate subsequent geographic information system (GIS) analysis and application.
[0060] In a possible embodiment of the present invention, the method further includes: constructing a training set for training a convolutional neural network, specifically:
[0061] Acquire the information of various disturbance spots collected on site, and establish an interpretation identification library after arranging and archiving the information of various disturbance spots, wherein the information of various disturbance spots includes specific location and type information;
[0062] Corresponding spots are drawn one by one on the collected remote sensing images according to the information in the interpretation identification library, and corresponding attribute information is added to each spot to form a training set, wherein the attribute information includes type, location, range size and description.
[0063] It should be noted that the information of various disturbance spots is sorted and archived to form a systematic interpretation library. The interpretation library contains the specific location, type and other relevant attribute information of each spot, and all spots and their attribute information are integrated to form a complete training set. The training set contains detailed information of various disturbance spots, providing rich sample data for the training of convolutional neural networks.
[0064] The second embodiment of the present invention provides a remote sensing image automatic interpretation device based on deep learning, comprising:
[0065] A preprocessing unit 201 is used to obtain multiple remote sensing images of the same sensor satellite in different regions and at different times, and to preprocess the remote sensing images;
[0066] The feature fusion unit 202 is used to call the pre-trained convolutional neural network to perform multi-layer convolution and pooling operations on the multi-period remote sensing images to generate feature maps at different levels, and fuse the feature maps at different levels into a comprehensive representation of the ground object target, wherein the feature map includes the edge, texture, shape and structure of the remote sensing image;
[0067] The classification unit 203 is used to call a pre-trained classifier to classify the comprehensive representation of the ground object target to generate a classification result, and integrate the classification result and output it in the form of a table, wherein the integration operation includes assigning different colors to each type of disturbance pattern of the classification result, marking the disturbance pattern with different shapes, and depicting the boundary of each disturbance pattern.
[0068] Preferably, the preprocessing includes radiation calibration, geometric correction, and atmospheric correction.
[0069] Preferably, the expression for generating feature maps at different levels is:
[0070]
[0071] in, It is a feature map In Location and The value on the convolution kernel, It is remote sensing image In Location and The pixel values on each channel are It is The convolution kernels are located at and The weights on the channels, , is the width and height of the convolution kernel, and It is the center position of the convolution kernel on the remote sensing image.
[0072] Preferably, the method further comprises: constructing a training set for training a convolutional neural network, specifically:
[0073] Obtain the information of various disturbance spots collected on site, and organize and archive the information of various disturbance spots to establish an interpretation identification library, wherein the information of various disturbance spots includes specific location and type information,
[0074] Corresponding spots are drawn one by one on the collected remote sensing images according to the information in the interpretation identification library, and corresponding attribute information is added to each spot to form a training set, wherein the attribute information includes type, location, range size and description.
[0075] The third embodiment of the present invention provides an automatic interpretation device for remote sensing images based on deep learning, including a memory and a processor, wherein the memory stores a computer program, and the computer program can be executed by the processor to implement an automatic interpretation method for remote sensing images based on deep learning as described in any one of the above.
[0076] The fourth embodiment of the present invention provides a computer-readable storage medium, characterized in that a computer program is stored therein, and the computer program can be executed by a processor of a device where the computer-readable storage medium is located to implement a method for automatic interpretation of remote sensing images based on deep learning as described in any one of the above items.
[0077] A method, device, equipment, and medium for automatic interpretation of remote sensing images based on deep learning provided by the present invention first obtain multi-period remote sensing images of the same sensor satellite in different regions and different periods, and pre-process the remote sensing images; then, call a pre-trained convolutional neural network to perform multi-layer convolution and pooling operations on the multi-period remote sensing images to generate feature maps of different levels, and fuse the feature maps of different levels into a comprehensive representation of ground objects; finally, call a pre-trained classifier to classify the comprehensive representation of the ground objects to generate a classification result, and integrate the classification results and output them in the form of a table, wherein the integration operation includes assigning different colors to each type of disturbance spots of the classification results, marking the disturbance spots with different shapes, and depicting the boundaries of each disturbance spot, thereby solving the problems of low efficiency and low accuracy of manual interpretation.
[0078] Exemplarily, the computer program described in the third and fourth embodiments of the present invention may be divided into one or more modules, and the one or more modules are stored in the memory and executed by the processor to complete the present invention. The one or more modules may be a series of computer program instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer program in the device for realizing an automatic interpretation of remote sensing images based on deep learning. For example, the device described in the second embodiment of the present invention.
[0079] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the automatic interpretation method of remote sensing images based on deep learning, and uses various interfaces and lines to connect the various parts of the automatic interpretation method of remote sensing images based on deep learning.
[0080] The memory can be used to store the computer program and / or module, and the processor realizes various functions of a remote sensing image automatic interpretation method based on deep learning by running or executing the computer program and / or module stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, a text conversion function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, text message data, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.
[0081] Wherein, if the implemented module is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.
[0082] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.
[0083] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.
Claims
1. A method for automatic interpretation of remote sensing images based on deep learning, characterized in that: include: Acquire multiple remote sensing images of the same sensor satellite in different regions and at different times, and preprocess the remote sensing images. The preprocessing includes radiation calibration and atmospheric correction. The radiation calibration is to convert the original image data into radiation brightness or reflectivity, thereby eliminating the difference between sensors and the influence of time changes. The atmospheric correction is to perform atmospheric correction on the image through the 6S model, combined with the metadata of the image and ground observation data, thereby removing the influence of the atmosphere on the remote sensing image. The pre-trained convolutional neural network is called to perform multi-layer convolution and pooling operations on the multi-period remote sensing images to generate feature maps at different levels, and the feature maps at different levels are fused into a comprehensive representation of the ground object target, specifically: In each layer of the convolutional neural network, the input image is scanned by multiple convolution kernels and downsampled by the pooling layer to generate feature maps at different levels. After the feature maps at different levels are fused, the linear transformation of the fully connected layer and the nonlinear transformation of the activation function are used to reduce the dimension and generate a comprehensive representation of the ground object target. The feature map includes the edge, texture, shape and structure of the remote sensing image; the expression for generating feature maps at different levels is: in, It is a feature map In Location and The value on the convolution kernel, It is remote sensing image In Location and The pixel values on each channel are It is The convolution kernels are located at and The weights on the channels, , is the width and height of the convolution kernel, and is the center position of the convolution kernel on the remote sensing image; A pre-trained classifier is called to classify the comprehensive representation of the ground object target to generate a classification result, and the classification result is integrated and output in a table form, wherein the integration operation includes assigning different colors to each type of disturbance pattern of the classification result, marking the disturbance pattern with different shapes, and depicting the boundary of each disturbance pattern.
2. The method for automatic interpretation of remote sensing images based on deep learning according to claim 1, characterized in that: Also includes: Construct a training set for training the convolutional neural network, specifically: Acquire the information of various disturbance spots collected on site, and establish an interpretation identification library after arranging and archiving the information of various disturbance spots, wherein the information of various disturbance spots includes specific location and type information; Corresponding spots are drawn one by one on the collected remote sensing images according to the information in the interpretation identification library, and corresponding attribute information is added to each spot to form a training set, wherein the attribute information includes type, location, range size and description.
3. A remote sensing image automatic interpretation device based on deep learning, characterized in that: include: A preprocessing unit is used to obtain multiple remote sensing images of the same sensor satellite in different regions and different periods, and preprocess the remote sensing images. The preprocessing includes radiation calibration and atmospheric correction. The radiation calibration is to convert the original image data into radiation brightness or reflectivity, so as to eliminate the difference between sensors and the influence of time changes. The atmospheric correction is to perform atmospheric correction on the image through the 6S model, combined with the metadata of the image and the ground observation data, so as to remove the influence of the atmosphere on the remote sensing image. The feature fusion unit is used to call the pre-trained convolutional neural network to perform multi-layer convolution and pooling operations on the multi-period remote sensing images to generate feature maps at different levels, and fuse the feature maps at different levels into a comprehensive representation of the ground object target, specifically used for: In each layer of the convolutional neural network, the input image is scanned by multiple convolution kernels and downsampled by the pooling layer to generate feature maps at different levels. After the feature maps at different levels are fused, the linear transformation of the fully connected layer and the nonlinear transformation of the activation function are used to reduce the dimension and generate a comprehensive representation of the ground object target. The feature map includes the edge, texture, shape and structure of the remote sensing image; the expression for generating feature maps at different levels is: in, It is a feature map In Location and The value on the convolution kernel, It is remote sensing image In Location and The pixel values on each channel are It is The convolution kernels are located at and The weights on the channels, , is the width and height of the convolution kernel, and is the center position of the convolution kernel on the remote sensing image; A classification unit is used to call a pre-trained classifier to classify the comprehensive representation of the ground object target to generate a classification result, and output the classification result in a table after integration operation, wherein the integration operation includes assigning different colors to each type of disturbance pattern of the classification result, marking the disturbance pattern with different shapes, and depicting the boundary of each disturbance pattern.
4. The deep learning-based remote sensing image automatic interpretation device according to claim 3, characterized in that: Also includes: Construct a training set for training the convolutional neural network, specifically: Obtain the information of various disturbance spots collected on site, and organize and archive the information of various disturbance spots to establish an interpretation identification library, wherein the information of various disturbance spots includes specific location and type information, Corresponding spots are drawn one by one on the collected remote sensing images according to the information in the interpretation identification library, and corresponding attribute information is added to each spot to form a training set, wherein the attribute information includes type, location, range size and description.
5. A remote sensing image automatic interpretation device based on deep learning, characterized in that: It includes a memory and a processor, wherein the memory stores a computer program, and the computer program can be executed by the processor to implement a method for automatic interpretation of remote sensing images based on deep learning as described in any one of claims 1 to 2.
6. A computer-readable storage medium, characterized in that: A computer program is stored, and the computer program can be executed by a processor of the device where the computer-readable storage medium is located to implement a method for automatic interpretation of remote sensing images based on deep learning as described in any one of claims 1 to 2.
Citation Information
Patent Citations
Remote sensing image scene classification method based on multi-scale depth feature fusion and transfer learning
CN110555446A
Remote sensing image water area automatic extraction method and system based on deep learning
CN111767801A
Improved high-resolution remote sensing image cultivated land extraction method based on deep learning
CN114926740A
Hyperspectral remote sensing image classification method based on attention joint network
CN115564996A
Attention enhancement and multi-scale feature fusion-based artificial disturbance ground remote sensing extraction method
CN116452991A
