Remote sensing image super-resolution reconstruction method, device and equipment based on artificial intelligence

By interpolation, preprocessing and model training of remote sensing images based on artificial intelligence, the problem of poor super-resolution reconstruction effect of remote sensing images in the prior art is solved, and high-quality image details recovery and higher reconstruction effect are achieved.

CN119067849APending Publication Date: 2024-12-03CHINA PING AN PROPERTY INSURANCE CO LTD
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411173512.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-23
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

The existing remote sensing image super-resolution reconstruction method based on interpolation method is prone to image blurring and lack of texture details, resulting in poor super-resolution reconstruction effect and difficult to meet the needs of high-precision applications.

Method used

Using an artificial intelligence-based method, the satellite remote sensing image pair is interpolated, preprocessed, cropped and aligned, and the remote sensing image samples are constructed, and the initial super-resolution reconstruction model is trained and evaluated to generate super-resolution that meets preset performance. Rate reconstruction model, used to perform super-resolution reconstruction of remote sensing images to be processed.

Benefits of technology

It realizes accurate recovery of the details of remote sensing images, outputs higher-quality super-resolution reconstruction results, and improves the super-resolution reconstruction effect of remote sensing images.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119067849A_ABST
    Figure CN119067849A_ABST
Patent Text Reader

Abstract

The invention belongs to the field of artificial intelligence and the field of financial science and technology, and relates to an artificial intelligence-based remote sensing image super-resolution reconstruction method, which comprises the following steps of: obtaining a satellite remote sensing image pair of a target area; performing interpolation on a second satellite remote sensing image in the satellite remote sensing image pair to obtain a third satellite remote sensing image; constructing a first satellite remote sensing image pair based on the first satellite remote sensing image and the third satellite remote sensing image; preprocessing the first satellite remote sensing image pair to obtain a remote sensing image sample; training based on a remote sensing image sample to obtain a super-resolution reconstruction model; and processing the remote sensing image based on the super-resolution reconstruction model to obtain a target remote sensing image. The invention further provides a remote sensing image super-resolution reconstruction device based on artificial intelligence, computer equipment and a storage medium. Based on the use of the super-resolution reconstruction model, the super-resolution reconstruction effect of the remote sensing image is effectively improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence development technology and financial technology, and in particular to artificial intelligence-based remote sensing image super-resolution reconstruction methods, devices, computer equipment and storage media. Background Art

[0002] In the wide application of remote sensing technology, remote sensing images, as an important data source, play an irreplaceable role in the fields of agricultural insurance, environmental monitoring, resource survey, etc. However, due to factors such as the resolution of remote sensing image sensors, optical system distortion, atmospheric interference, and imaging system noise, the acquired remote sensing images often have insufficient resolution, which greatly limits their effectiveness in high-precision applications, such as accurate loss assessment and risk assessment in agricultural insurance.

[0003] At present, in the field of agricultural insurance, the super-resolution reconstruction method of remote sensing images mainly adopts interpolation method. Although interpolation method can improve the resolution of remote sensing images to a certain extent, its effect is often limited and it is difficult to meet the needs of high-precision applications. Interpolation methods such as bilinear interpolation and bicubic interpolation are simple to calculate, but they are prone to image blur and lack of texture details, resulting in poor super-resolution reconstruction effect of remote sensing images. Summary of the invention

[0004] The purpose of the embodiments of the present application is to propose a remote sensing image super-resolution reconstruction method, device, computer equipment and storage medium based on artificial intelligence to solve the technical problem that the existing remote sensing image super-resolution reconstruction method based on interpolation method, although simple in calculation, is prone to image blur and lack of texture details, resulting in poor super-resolution reconstruction effect of remote sensing images.

[0005] In order to solve the above technical problems, the present application embodiment provides a remote sensing image super-resolution reconstruction method based on artificial intelligence, which adopts the following technical solutions:

[0006] Acquire a pair of pre-collected satellite remote sensing images of a target area; wherein the pair of satellite remote sensing images includes a first satellite remote sensing image having a first resolution and a second satellite remote sensing image having a second resolution, and the first resolution is greater than the second resolution;

[0007] Performing interpolation processing on the second satellite remote sensing image to obtain a corresponding third satellite remote sensing image; wherein the resolution of the third satellite remote sensing image is a third resolution, and the third resolution is greater than the second resolution;

[0008] Constructing a corresponding first satellite remote sensing image pair based on the first satellite remote sensing image and the third satellite remote sensing image;

[0009] Preprocessing the first satellite remote sensing image pair to obtain corresponding remote sensing image samples;

[0010] Based on the remote sensing image samples, a preset initial super-resolution reconstruction model is trained and evaluated to obtain a super-resolution reconstruction model that meets preset performance expectations;

[0011] The remote sensing image to be processed is subjected to super-resolution reconstruction processing based on the super-resolution reconstruction model to obtain a corresponding target remote sensing image.

[0012] Furthermore, the step of performing interpolation processing on the second satellite remote sensing image to obtain a corresponding third satellite remote sensing image specifically includes:

[0013] Get the target interpolation method;

[0014] Obtaining a preset target resolution; wherein the target resolution is equal to the third resolution;

[0015] performing interpolation processing on the second satellite remote sensing image based on the target interpolation method and the target resolution to obtain a processed second satellite remote sensing image;

[0016] The processed second satellite remote sensing image is used as the third satellite remote sensing image.

[0017] Furthermore, the step of preprocessing the first satellite remote sensing image pair to obtain corresponding remote sensing image samples specifically includes:

[0018] Get the preset image processing strategy;

[0019] Processing the first satellite remote sensing image pair based on the image processing strategy to obtain a corresponding second satellite remote sensing image pair;

[0020] Get preset ground truth high-resolution images;

[0021] Performing registration processing on the second satellite remote sensing image pair based on a preset registration strategy and the ground true high-resolution image to obtain a corresponding third satellite remote sensing image pair;

[0022] Performing cropping and alignment processing on the third satellite remote sensing image pair to obtain a corresponding fourth satellite remote sensing image pair;

[0023] The fourth satellite remote sensing image pair is used as the remote sensing image sample.

[0024] Furthermore, the step of performing cropping and alignment processing on the third satellite remote sensing image pair to obtain a corresponding fourth satellite remote sensing image pair specifically includes:

[0025] Obtaining preset slicing values; wherein the slicing values ​​include a first slicing value corresponding to a high-resolution image and a second slicing value corresponding to a low-resolution image;

[0026] Performing a cropping process on the third satellite remote sensing image pair based on the block value to obtain a corresponding first image pair;

[0027] Performing image spatial domain alignment processing on the first image pair to obtain a corresponding second image pair;

[0028] The second image pair is used as the fourth satellite remote sensing image pair.

[0029] Furthermore, the initial super-resolution reconstruction model includes a generator and a discriminator, the generator includes a convolution layer, a target number of feature extraction layers and an upsampling layer connected in sequence, and the discriminator adopts a spectrally normalized Unet structure; before the step of training and evaluating the preset initial super-resolution reconstruction model based on the remote sensing image sample to obtain a super-resolution reconstruction model that meets the preset performance expectations, it also includes:

[0030] Dividing the remote sensing image samples into a training data set and a test data set according to a preset ratio;

[0031] Calling the initial super-resolution reconstruction model;

[0032] Get the preset model training strategy;

[0033] Based on the model training strategy, the initial super-resolution reconstruction model is trained using the training data set to obtain a corresponding designated model;

[0034] Performing a performance evaluation on the specified model based on the test data set;

[0035] If the specified model passes the performance evaluation, the specified model is used as the super-resolution reconstruction model.

[0036] Furthermore, the step of obtaining a preset model training strategy specifically includes:

[0037] Get target optimizer information;

[0038] Get the preset learning rate adjustment strategy;

[0039] Get the target loss function;

[0040] The model training strategy is generated based on the target optimizer information, the learning rate adjustment strategy, and the target loss function.

[0041] Furthermore, after the step of training and evaluating the preset initial super-resolution reconstruction model based on the remote sensing image samples to obtain a super-resolution reconstruction model that meets the preset performance expectations, the method further includes:

[0042] Get the preset model deployment format;

[0043] Exporting the super-resolution reconstruction model based on the model deployment format to obtain a target super-resolution reconstruction model;

[0044] The target super-resolution reconstruction model is stored and integrated.

[0045] In order to solve the above technical problems, the embodiment of the present application also provides a remote sensing image super-resolution reconstruction device based on artificial intelligence, which adopts the following technical solution:

[0046] A first acquisition module is used to acquire a pair of satellite remote sensing images of a target area collected in advance; wherein the pair of satellite remote sensing images includes a first satellite remote sensing image with a first resolution and a second satellite remote sensing image with a second resolution, and the first resolution is greater than the second resolution;

[0047] An interpolation module, configured to perform interpolation processing on the second satellite remote sensing image to obtain a corresponding third satellite remote sensing image; wherein the resolution of the third satellite remote sensing image is a third resolution, and the third resolution is greater than the second resolution;

[0048] A construction module, configured to construct a corresponding first satellite remote sensing image pair based on the first satellite remote sensing image and the third satellite remote sensing image;

[0049] A preprocessing module, used for preprocessing the first satellite remote sensing image pair to obtain corresponding remote sensing image samples;

[0050] A training module, used to train and evaluate a preset initial super-resolution reconstruction model based on the remote sensing image samples to obtain a super-resolution reconstruction model that meets preset performance expectations;

[0051] The reconstruction module is used to perform super-resolution reconstruction processing on the remote sensing image to be processed based on the super-resolution reconstruction model to obtain the corresponding target remote sensing image.

[0052] In order to solve the above technical problems, the embodiment of the present application further provides a computer device, which adopts the following technical solution:

[0053] Acquire a pair of pre-collected satellite remote sensing images of a target area; wherein the pair of satellite remote sensing images includes a first satellite remote sensing image having a first resolution and a second satellite remote sensing image having a second resolution, and the first resolution is greater than the second resolution;

[0054] Performing interpolation processing on the second satellite remote sensing image to obtain a corresponding third satellite remote sensing image; wherein the resolution of the third satellite remote sensing image is a third resolution, and the third resolution is greater than the second resolution;

[0055] Constructing a corresponding first satellite remote sensing image pair based on the first satellite remote sensing image and the third satellite remote sensing image;

[0056] Preprocessing the first satellite remote sensing image pair to obtain corresponding remote sensing image samples;

[0057] Based on the remote sensing image samples, a preset initial super-resolution reconstruction model is trained and evaluated to obtain a super-resolution reconstruction model that meets preset performance expectations;

[0058] The remote sensing image to be processed is subjected to super-resolution reconstruction processing based on the super-resolution reconstruction model to obtain a corresponding target remote sensing image.

[0059] In order to solve the above technical problems, the embodiment of the present application further provides a computer-readable storage medium, which adopts the following technical solution:

[0060] Acquire a pair of pre-collected satellite remote sensing images of a target area; wherein the pair of satellite remote sensing images includes a first satellite remote sensing image having a first resolution and a second satellite remote sensing image having a second resolution, and the first resolution is greater than the second resolution;

[0061] Performing interpolation processing on the second satellite remote sensing image to obtain a corresponding third satellite remote sensing image; wherein the resolution of the third satellite remote sensing image is a third resolution, and the third resolution is greater than the second resolution;

[0062] Constructing a corresponding first satellite remote sensing image pair based on the first satellite remote sensing image and the third satellite remote sensing image;

[0063] Preprocessing the first satellite remote sensing image pair to obtain corresponding remote sensing image samples;

[0064] Based on the remote sensing image samples, a preset initial super-resolution reconstruction model is trained and evaluated to obtain a super-resolution reconstruction model that meets preset performance expectations;

[0065] The remote sensing image to be processed is subjected to super-resolution reconstruction processing based on the super-resolution reconstruction model to obtain a corresponding target remote sensing image.

[0066] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0067] This application first obtains a pair of pre-collected satellite remote sensing images of the target area; then interpolates the second satellite remote sensing image to obtain a corresponding third satellite remote sensing image; then constructs a corresponding first satellite remote sensing image pair based on the first satellite remote sensing image and the third satellite remote sensing image; and preprocesses the first satellite remote sensing image pair to obtain a corresponding remote sensing image sample; further trains and evaluates a preset initial super-resolution reconstruction model based on the remote sensing image sample to obtain a super-resolution reconstruction model that meets preset performance expectations; finally, performs super-resolution reconstruction processing on the remote sensing image to be processed based on the super-resolution reconstruction model to obtain the corresponding target remote sensing image. This application interpolates, preprocesses, crops and aligns pre-collected satellite remote sensing image pairs of the target area to obtain remote sensing image samples, and then trains and evaluates a preset initial super-resolution reconstruction model based on the remote sensing image samples to obtain a super-resolution reconstruction model that meets preset performance expectations, and subsequently performs super-resolution reconstruction processing on the remote sensing image to be processed based on the super-resolution reconstruction model to obtain the corresponding target remote sensing image. This application uses a super-resolution reconstruction model built based on remote sensing image samples to perform super-resolution reconstruction processing of remote sensing images. It can accurately restore the detailed texture information of the remote sensing image to be processed and output higher quality super-resolution reconstruction results, effectively improving the super-resolution reconstruction effect of remote sensing images. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] In order to more clearly illustrate the scheme in the present application, a brief introduction is given below to the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0069] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;

[0070] Figure 2 A flowchart of an embodiment of a remote sensing image super-resolution reconstruction method based on artificial intelligence according to the present application;

[0071] Figure 3 It is a structural schematic diagram of an embodiment of a remote sensing image super-resolution reconstruction device based on artificial intelligence according to the present application;

[0072] Figure 4 It is a structural diagram of an embodiment of a computer device according to the present application. DETAILED DESCRIPTION

[0073] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0074] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0075] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0076] like Figure 1 As shown, the system architecture 100 may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0077] Users can use terminal devices 101, 102, 103 to interact with server 105 through network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, 103, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0078] Terminal devices 101, 102, 103 can be various electronic devices with display screens and supporting web browsing, including but not limited to smart phones, tablet computers, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III), MP4 (Moving Picture Experts Group Audio Layer IV) players, laptop computers, desktop computers, etc.

[0079] The server 105 may be a server that provides various services, such as a background server that provides support for web pages displayed on the terminal devices 101 , 102 , and 103 .

[0080] It should be noted that the artificial intelligence-based remote sensing image super-resolution reconstruction method provided in the embodiment of the present application is generally executed by a server / terminal device, and accordingly, the artificial intelligence-based remote sensing image super-resolution reconstruction device is generally set in the server / terminal device.

[0081] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to the implementation requirements.

[0082] Continue to refer Figure 2 , showing a flowchart of an embodiment of the artificial intelligence-based remote sensing image super-resolution reconstruction method according to the present application. According to different needs, the order of the steps in the flowchart can be changed, and some steps can be omitted. The artificial intelligence-based remote sensing image super-resolution reconstruction method provided in the embodiment of the present application can be applied to any scenario that requires super-resolution reconstruction of low-resolution remote sensing images, and the artificial intelligence-based remote sensing image super-resolution reconstruction method can be applied to products in these scenarios, for example, the detail restoration of low-resolution remote sensing images in the field of agricultural insurance applications. The artificial intelligence-based remote sensing image super-resolution reconstruction method comprises the following steps:

[0083] Step S201, obtaining a pair of pre-collected satellite remote sensing images of a target area; wherein the pair of satellite remote sensing images includes a first satellite remote sensing image with a first resolution and a second satellite remote sensing image with a second resolution, and the first resolution is greater than the second resolution.

[0084] In this embodiment, the remote sensing image super-resolution reconstruction method based on artificial intelligence is run on the electronic device (for example Figure 1The server / terminal device shown in the figure) can obtain the satellite remote sensing image pair of the target area through a wired connection or a wireless connection. It should be noted that the above-mentioned wireless connection method may include but is not limited to 3G / 4G / 5G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other wireless connection methods currently known or to be developed in the future. The above-mentioned target area may be an area where claims processing is required in agricultural insurance business. Among them, the above-mentioned first resolution may be specifically 10 meters, and the above-mentioned second resolution may be specifically 2 meters. By collecting satellite remote sensing images with a resolution of 10 meters and a resolution of 2 meters of the same target area to obtain the above-mentioned satellite remote sensing image pair, there is a mapping relationship between the first satellite remote sensing image with a resolution of 10 meters and the second satellite remote sensing image with a resolution of 2 meters. In addition, during the acquisition of satellite remote sensing images, ensure that the shooting time of satellite remote sensing images with a resolution of 10 meters and a resolution of 2 meters is similar to reduce the image differences caused by seasonal changes, changes in surface cover and other factors. In addition, it is also necessary to ensure that the satellite remote sensing images are free of cloud and fog to ensure the quality of the satellite remote sensing images.

[0085] Step S202: interpolate the second satellite remote sensing image to obtain a corresponding third satellite remote sensing image; wherein the resolution of the third satellite remote sensing image is a third resolution, and the third resolution is greater than the second resolution.

[0086] In this embodiment, the third resolution may be 2.5 meters. The specific implementation process of interpolating the second satellite remote sensing image to obtain the corresponding third satellite remote sensing image will be further described in detail in subsequent specific embodiments of the present application, and will not be elaborated here.

[0087] Step S203: construct a corresponding first satellite remote sensing image pair based on the first satellite remote sensing image and the third satellite remote sensing image.

[0088] In this embodiment, by using the first satellite remote sensing image as a low-resolution image and the third satellite remote sensing image as a high-resolution image, the first satellite remote sensing image pair constructed includes a high-resolution third satellite remote sensing image having a mapping relationship and a low-resolution first satellite remote sensing image.

[0089] Step S204: preprocess the first satellite remote sensing image pair to obtain corresponding remote sensing image samples.

[0090] In this embodiment, the above-mentioned specific implementation process of preprocessing the first satellite remote sensing image pair to obtain the corresponding remote sensing image samples will be further described in detail in subsequent specific embodiments of the present application and will not be elaborated on here.

[0091] Step S205 , training and evaluating a preset initial super-resolution reconstruction model based on the remote sensing image samples to obtain a super-resolution reconstruction model that meets preset performance expectations.

[0092] In this embodiment, the above-mentioned specific implementation process of training and evaluating the preset initial super-resolution reconstruction model based on the remote sensing image samples to obtain a super-resolution reconstruction model that meets the preset performance expectations will be further described in detail in the subsequent specific embodiments of the present application and will not be elaborated in detail here.

[0093] Step S206, performing super-resolution reconstruction processing on the remote sensing image to be processed based on the super-resolution reconstruction model to obtain a corresponding target remote sensing image.

[0094] In this embodiment, the remote sensing image to be processed is specifically a low-resolution remote sensing image. By inputting the remote sensing image to be processed into the super-resolution reconstruction model, the super-resolution reconstruction model performs super-resolution reconstruction processing on the remote sensing image to be processed, so as to accurately restore the detail texture information of the remote sensing image to be processed, and output a higher quality super-resolution reconstruction result, that is, the target remote sensing image.

[0095] This application first obtains a pair of pre-collected satellite remote sensing images of the target area; then interpolates the second satellite remote sensing image to obtain a corresponding third satellite remote sensing image; then constructs a corresponding first satellite remote sensing image pair based on the first satellite remote sensing image and the third satellite remote sensing image; and preprocesses the first satellite remote sensing image pair to obtain a corresponding remote sensing image sample; further trains and evaluates a preset initial super-resolution reconstruction model based on the remote sensing image sample to obtain a super-resolution reconstruction model that meets preset performance expectations; finally, performs super-resolution reconstruction processing on the remote sensing image to be processed based on the super-resolution reconstruction model to obtain the corresponding target remote sensing image. This application interpolates, preprocesses, crops and aligns pre-collected satellite remote sensing image pairs of the target area to obtain remote sensing image samples, and then trains and evaluates a preset initial super-resolution reconstruction model based on the remote sensing image samples to obtain a super-resolution reconstruction model that meets preset performance expectations, and subsequently performs super-resolution reconstruction processing on the remote sensing image to be processed based on the super-resolution reconstruction model to obtain the corresponding target remote sensing image. This application uses a super-resolution reconstruction model built based on remote sensing image samples to perform super-resolution reconstruction processing of remote sensing images. It can accurately restore the detailed texture information of the remote sensing image to be processed and output higher quality super-resolution reconstruction results, effectively improving the super-resolution reconstruction effect of remote sensing images.

[0096] In some optional implementations of this embodiment, step S202 includes the following steps:

[0097] Get the target interpolation method.

[0098] In this embodiment, the target interpolation method may specifically adopt a bilinear interpolation method.

[0099] Obtain a preset target resolution; wherein the target resolution is equal to the third resolution.

[0100] In this embodiment, the target resolution is specifically the third resolution, i.e., 2.5 meters. Since the resolution multiple corresponding to the 2-meter resolution data is not 2 to the power of 2, the second satellite remote sensing image with a resolution of 2 meters needs to be interpolated to a resolution of 2.5 meters.

[0101] The second satellite remote sensing image is interpolated based on the target interpolation method and the target resolution to obtain a processed second satellite remote sensing image.

[0102] In this embodiment, the second satellite remote sensing image is interpolated by using the target interpolation method so that the resolution of the interpolated second satellite remote sensing image is the target resolution.

[0103] The processed second satellite remote sensing image is used as the third satellite remote sensing image.

[0104] The present application obtains a target interpolation method; then obtains a target resolution; wherein the target resolution is equal to a third resolution; and then interpolates the second satellite remote sensing image based on the target interpolation method and the target resolution to obtain the third satellite remote sensing image that meets the resolution requirements, thereby improving the generation efficiency and generation intelligence of the third satellite remote sensing image and ensuring the accuracy of the resolution value of the obtained third satellite remote sensing image.

[0105] In some optional implementations, step S204 includes the following steps:

[0106] Get the preset image processing strategy.

[0107] In this embodiment, the above-mentioned image processing strategies specifically include strategies of image fusion, geographic registration, radiation calibration, atmospheric correction, and image stretching processing.

[0108] The first satellite remote sensing image pair is processed based on the image processing strategy to obtain a corresponding second satellite remote sensing image pair.

[0109] In this embodiment, based on the image processing strategy, the first satellite remote sensing image pair is subjected to image fusion, geographic registration, radiation calibration, atmospheric correction and image stretching processing to obtain a corresponding second satellite remote sensing image pair, thereby effectively ensuring the image accuracy and consistency of the second satellite remote sensing image pair.

[0110] Get preset ground truth high-resolution imagery.

[0111] In this embodiment, the above-mentioned ground true high-resolution image may specifically refer to the ground true high-resolution image of the above-mentioned target area collected by a drone, or may also be referred to as a drone image matched with the above-mentioned satellite remote sensing image pair.

[0112] The second satellite remote sensing image pair is registered based on a preset registration strategy and the ground true high-resolution image to obtain a corresponding third satellite remote sensing image pair.

[0113] In this embodiment, the ground true high-resolution image and the second satellite remote sensing image pair can be registered by adopting a feature point-based registration algorithm or a grayscale-based registration algorithm, and the accuracy can be controlled within 0.5 pixels to obtain a third satellite remote sensing image pair. This ensures that the satellite remote sensing image and the ground true high-resolution image are accurately aligned in space, reduces errors caused by position deviation, and effectively improves the data accuracy of the third satellite remote sensing image pair.

[0114] Performing cropping and alignment processing on the third satellite remote sensing image pair to obtain a corresponding fourth satellite remote sensing image pair;

[0115] In this embodiment, the above-mentioned cropping and alignment processing of the third satellite remote sensing image pair to obtain the corresponding fourth satellite remote sensing image pair is specifically implemented. This application will provide further details on this in subsequent specific embodiments and will not be elaborated on here.

[0116] The fourth satellite remote sensing image pair is used as the remote sensing image sample.

[0117] The present application processes the first satellite remote sensing image pair based on a preset image processing strategy to obtain a corresponding second satellite remote sensing image pair, then performs registration processing on the second satellite remote sensing image pair based on a preset registration strategy and the acquired ground true high-resolution image to obtain a corresponding third satellite remote sensing image pair, and subsequently performs cropping and alignment processing on the third satellite remote sensing image pair to obtain corresponding remote sensing image samples, thereby achieving fast and intelligent preprocessing of the first satellite remote sensing image pair and effectively improving the data accuracy of the generated remote sensing image samples.

[0118] In some optional implementations, the cropping and alignment processing of the third satellite remote sensing image pair to obtain a corresponding fourth satellite remote sensing image pair includes the following steps:

[0119] Obtaining preset slicing values; wherein the slicing values ​​include a first slicing value corresponding to a high-resolution image, and a second slicing value corresponding to a low-resolution image.

[0120] In this embodiment, since the remote sensing image is too large to be directly fed into the model for training, it needs to be cropped into small slices. The above-mentioned first slice value and the second slice value are pre-built slice values ​​that meet the model input requirements. Exemplarily, the above-mentioned first slice value corresponding to the high-resolution image can be specifically set to 512×512, and it is cropped every 512 pixels without overlapping. The above-mentioned second slice value corresponding to the low-resolution image can be specifically set to 128*128, and it is cropped every 128 pixels without overlapping.

[0121] The third satellite remote sensing image pair is cropped based on the block value to obtain a corresponding first image pair.

[0122] In this embodiment, the third satellite remote sensing image pair is cropped by using the block value, thereby obtaining a first image pair of high and low resolutions whose pixel sizes are the first block value and the second block value.

[0123] Performing image spatial domain alignment processing on the first image pair to obtain a corresponding second image pair.

[0124] In this embodiment, by performing image spatial alignment processing on the first image pair, the image difference caused by the difference in shooting time can be effectively reduced, and the distribution of the two high and low resolution image data in the first image pair can be made as consistent as possible, thereby obtaining the corresponding second image pair.

[0125] The second image pair is used as the fourth satellite remote sensing image pair.

[0126] The present application uses the acquired block value to perform cropping processing on the third satellite remote sensing image pair to obtain the corresponding first image pair, and then performs image spatial domain alignment processing on the first image pair, so as to quickly and accurately obtain the fourth satellite remote sensing image pair, so that the fourth satellite remote sensing image pair is subsequently used as the final remote sensing image sample, effectively improving the construction efficiency of the remote sensing image sample and improving the data accuracy of the generated remote sensing image sample.

[0127] In some optional implementations, the initial super-resolution reconstruction model includes a generator and a discriminator, the generator includes a convolution layer, a target number of feature extraction layers and an upsampling layer connected in sequence, and the discriminator adopts a spectrally normalized Unet structure; step S205 includes the following steps:

[0128] The remote sensing image samples are divided into a training data set and a test data set according to a preset ratio.

[0129] In this embodiment, there is no specific limitation on the numerical value selection of the above preset ratio, which can be set according to the actual model training requirements, for example, it can be set to 7:3.

[0130] The initial super-resolution reconstruction model is called.

[0131] In this embodiment, the initial super-resolution reconstruction model is a model based on Real-ESRGAN (Training Real-World Blind Super-Resolution with Pure Synthetic Data, using pure synthetic training to train the real-world blind super-resolution model). The Real-ESRGAN model can learn complex ground objects and scene features, so that it has better adaptability and generalization ability when processing complex scenes. Specifically, the initial super-resolution reconstruction model includes a generator and a discriminator. The above-mentioned target number can be specifically 23. The above-mentioned generator includes a convolutional layer for preliminary feature extraction and 23 Residual in Residual DenseBlocks (RRDB) feature extraction layers for deep feature extraction, each RDDB block is composed of 3 ResidualDenseBlocks, and the input and output shapes are consistent. It also includes an upsampling layer, which is used to perform two upsampling operations, using the nearest interpolation method, and after each upsampling, there will be a convolutional layer to refine the interpolation details. Among them, the RRDB block removes Batch Normlization, which can reduce the computational complexity and potential problems during model training. The above discriminator is used to calculate the feature difference between the generated remote sensing image data and the real drone image data. The discriminator adopts the Unet structure of spectral normalization, including DownSample and UpSample parts, with a lateral connection (Skip Connection) in the middle to enhance feature transfer. By using Unet as a discriminator, the ability of the discriminator and the stability of model training can be effectively increased.

[0132] Get the preset model training strategy.

[0133] In this embodiment, the specific implementation process of obtaining the preset model training strategy mentioned above will be further described in detail in subsequent specific embodiments of the present application and will not be elaborated in detail here.

[0134] Based on the model training strategy, the initial super-resolution reconstruction model is trained using the training data set to obtain a corresponding designated model.

[0135] In this embodiment, the model training strategy includes target optimizer information, learning rate adjustment strategy and target loss function. The training data set is input into the initial super-resolution reconstruction model for forward propagation to obtain the prediction result, and then the loss between the model prediction and the real label is calculated according to the target loss function, and then the optimizer corresponding to the target optimizer information is used for back propagation to update the model weight, and the learning rate scheduler corresponding to the learning rate adjustment strategy is applied to update the learning rate. Repeat the above training steps until the predetermined training rounds are reached or other stop conditions are met, so as to obtain a trained specified model. Among them, the target loss function includes a generator loss function and a discriminator loss function. During the training process of the initial super-resolution reconstruction model, the generator and the discriminator compete by alternately optimizing their loss functions. The generator tries to minimize its loss, while the discriminator tries to maximize its loss. This adversarial training process gradually enables the generator to generate super-resolution images that are closer and closer to the real image. In addition, the model weights and optimizer states are regularly saved during the model training process to resume training or evaluation later.

[0136] A performance evaluation is performed on the specified model based on the test data set.

[0137] In this embodiment, the above-mentioned test data set is input into the specified model to evaluate the model performance of the third specified model using the test data set, and the corresponding performance evaluation index is calculated. Then, it is determined whether the performance evaluation index is greater than the preset index threshold. If the evaluation index is greater than the index threshold, the specified model is determined to have passed the performance evaluation, otherwise the specified model is determined to have failed the performance evaluation. Among them, the above-mentioned evaluation index may include PSNR or SSIM index. PSNR is an index for measuring image quality, which is calculated by comparing the difference between the original image and the distorted image. Specifically, PSNR is calculated by comparing each pixel value of the two images. SSIM (structural similarity) is an index for measuring the similarity of two images. In addition, the selection of the above-mentioned index threshold can be set according to actual business usage requirements.

[0138] If the specified model passes the performance evaluation, the specified model is used as the super-resolution reconstruction model.

[0139] In this embodiment, if the specified model passes the performance evaluation, it indicates that the specified model is a model that meets the performance expectations, and the specified model will be used as the super-resolution reconstruction model. In addition, if the specified model fails the performance evaluation, the model parameters of the specified model can be further adjusted according to the test results of the performance evaluation and the model can be retrained until a model that meets the performance expectations is obtained and used as the super-resolution reconstruction model.

[0140] The present application divides the remote sensing image samples into a training data set and a test data set according to a preset ratio, and then, based on the acquired model training strategy, uses the training data set to train the initial super-resolution reconstruction model to obtain a corresponding designated model, and performs performance evaluation on the designated model based on the test data set, and when it is detected that the designated model passes the performance evaluation, uses the designated model as the super-resolution reconstruction model, so as to quickly and intelligently build a super-resolution reconstruction model that meets performance expectations, and effectively ensures the model effect of the generated super-resolution reconstruction model.

[0141] In some optional implementations of this embodiment, the step of obtaining a preset model training strategy includes the following steps:

[0142] Get target optimizer information.

[0143] In this embodiment, the target optimizer information may specifically use Adam, which performs well in super-resolution reconstruction tasks due to its adaptive learning rate adjustment strategy.

[0144] Get the preset learning rate adjustment strategy.

[0145] In this embodiment, the learning rate adjustment strategy includes: setting the initial learning rate to a smaller value (such as 1e-4) and using a cosine annealing learning rate scheduler to adjust the learning rate during training to improve training efficiency and stability.

[0146] Get the target loss function.

[0147] In this embodiment, the above-mentioned target loss function includes a generator loss function and a discriminator loss function. The choice of the target loss function can be weighed according to the image reconstruction requirements and the characteristics of the data set. Specifically, the goal of the generator is to generate a super-resolution image that is as close to the real high-resolution image as possible. The corresponding generator loss function includes multiple parts to evaluate the quality of the generated image from different perspectives. The generator loss includes content loss, perceptual loss and adversarial loss. Among them, the content loss specifically adopts the mean absolute error to calculate the sum of the absolute differences at the pixel level between the generated image and the real image, which encourages the generated image to be similar to the real image at the pixel level. Perceptual loss is based on a pre-trained network to evaluate the difference between the generated image and the real image in high-level feature representation. It works by comparing the feature maps of the two images in the pre-trained network, encouraging the generated image to be closer to the real image semantically and visually. The adversarial loss is provided by the discriminator to evaluate whether the generated image is real enough to deceive the discriminator. The generator tries to minimize the adversarial loss so that the discriminator cannot distinguish between its generated image and the real high-resolution image. In addition, the goal of the discriminator is to distinguish whether the input image is real or generated by the generator. The loss function is usually designed to maximize the accuracy of the correct classification (real or fake). Specifically, the binary cross entropy loss can be used as the discriminator loss function. For each input image, the discriminator outputs a score between 0 and 1, indicating the probability that the image is real. Then, the binary cross entropy loss function is used to calculate the difference between the predicted probability and the true label (0 for fake and 1 for real).

[0148] The model training strategy is generated based on the target optimizer information, the learning rate adjustment strategy, and the target loss function.

[0149] In this embodiment, a corresponding model training strategy can be obtained by formulating a training strategy for the target optimizer information, the learning rate adjustment strategy, and the target loss function.

[0150] The present application obtains target optimizer information, a preset learning rate adjustment strategy, and a target loss function, and then generates the model training strategy based on the target optimizer information, the learning rate adjustment strategy, and the target loss function, so as to quickly and intelligently generate a model training strategy that meets the requirements, which is beneficial for subsequent training of the initial super-resolution reconstruction model based on the model training strategy using the training data set, and can quickly and accurately obtain a trained super-resolution reconstruction model.

[0151] In some optional implementations of this embodiment, after step S205, the electronic device may further perform the following steps:

[0152] Get the preset model deployment format.

[0153] In this embodiment, there is no specific limitation on the selection of the above-mentioned model deployment format, which can be set according to the actual model deployment requirements, for example, it can include deployable formats such as .pth, .h5, etc.

[0154] The super-resolution reconstruction model is exported based on the model deployment format to obtain a target super-resolution reconstruction model.

[0155] In this embodiment, the target super-resolution reconstruction model is obtained by exporting the super-resolution reconstruction model into a format corresponding to the model deployment format.

[0156] The target super-resolution reconstruction model is stored and integrated.

[0157] In this embodiment, the target super-resolution reconstruction model can be stored by using a preset storage method. The above storage method is not specifically limited, for example, blockchain storage, local database storage, cloud server storage, etc. can be used. In addition, the target super-resolution reconstruction model can be integrated into actual application scenarios, such as image processing software, mobile applications, etc., to complete the integration of the target super-resolution reconstruction model.

[0158] After training and evaluating the preset initial super-resolution reconstruction model based on the remote sensing image samples and obtaining a super-resolution reconstruction model that meets the preset performance expectations, the application will also intelligently export the super-resolution reconstruction model based on the preset model deployment format to obtain a target super-resolution reconstruction model, and then store the target super-resolution reconstruction model to ensure the security of the generated target super-resolution reconstruction model. In addition, the target super-resolution reconstruction model will also be integrated so that the target super-resolution reconstruction model can be successfully applied in actual application scenarios.

[0159] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0160] It should be emphasized that in order to further ensure the privacy and security of the above-mentioned target remote sensing images, the above-mentioned target remote sensing images can also be stored in a blockchain node.

[0161] The blockchain referred to in this application is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0162] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0163] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0164] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through computer-readable instructions, and the computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0165] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0166] Further references Figure 3 , as a response to the above Figure 2 The present application provides an embodiment of a remote sensing image super-resolution reconstruction device based on artificial intelligence, and the device embodiment is similar to Figure 2 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0167] like Figure 3 As shown, the artificial intelligence-based remote sensing image super-resolution reconstruction device 300 described in this embodiment includes: a first acquisition module 301, an interpolation module 302, a construction module 303, a preprocessing module 304, a training module 305 and a reconstruction module 306. Among them:

[0168] The first acquisition module 301 is used to acquire a pair of satellite remote sensing images of a target area collected in advance; wherein the pair of satellite remote sensing images includes a first satellite remote sensing image with a first resolution and a second satellite remote sensing image with a second resolution, and the first resolution is greater than the second resolution;

[0169] An interpolation module 302 is used to perform interpolation processing on the second satellite remote sensing image to obtain a corresponding third satellite remote sensing image; wherein the resolution of the third satellite remote sensing image is a third resolution, and the third resolution is greater than the second resolution;

[0170] A construction module 303 is used to construct a corresponding first satellite remote sensing image pair based on the first satellite remote sensing image and the third satellite remote sensing image;

[0171] A preprocessing module 304 is used to preprocess the first satellite remote sensing image pair to obtain corresponding remote sensing image samples;

[0172] A training module 305 is used to train and evaluate a preset initial super-resolution reconstruction model based on the remote sensing image samples to obtain a super-resolution reconstruction model that meets preset performance expectations;

[0173] The reconstruction module 306 is used to perform super-resolution reconstruction processing on the remote sensing image to be processed based on the super-resolution reconstruction model to obtain a corresponding target remote sensing image.

[0174] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based remote sensing image super-resolution reconstruction method in the aforementioned implementation, and will not be repeated here.

[0175] In some optional implementations of this embodiment, the interpolation module 302 includes:

[0176] A first acquisition submodule is used to acquire a target interpolation method;

[0177] A second acquisition submodule is used to acquire a preset target resolution; wherein the target resolution is equal to the third resolution;

[0178] An interpolation submodule, configured to perform interpolation processing on the second satellite remote sensing image based on the target interpolation mode and the target resolution to obtain a processed second satellite remote sensing image;

[0179] The first determining submodule is used to use the processed second satellite remote sensing image as the third satellite remote sensing image.

[0180] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based remote sensing image super-resolution reconstruction method in the aforementioned implementation, and will not be repeated here.

[0181] In some optional implementations of this embodiment, the preprocessing module 304 includes:

[0182] The third acquisition submodule is used to acquire a preset image processing strategy;

[0183] A first processing submodule, configured to process the first satellite remote sensing image pair based on the image processing strategy to obtain a corresponding second satellite remote sensing image pair;

[0184] The fourth acquisition submodule is used to acquire a preset ground truth high-resolution image;

[0185] A registration submodule, configured to perform registration processing on the second satellite remote sensing image pair based on a preset registration strategy and the ground true high-resolution image to obtain a corresponding third satellite remote sensing image pair;

[0186] A second processing submodule is used to perform cropping and alignment processing on the third satellite remote sensing image pair to obtain a corresponding fourth satellite remote sensing image pair;

[0187] The second determining submodule is used to use the fourth satellite remote sensing image pair as the remote sensing image sample.

[0188] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based remote sensing image super-resolution reconstruction method in the aforementioned implementation, and will not be repeated here.

[0189] In some optional implementations of this embodiment, the second processing submodule includes:

[0190] A first acquisition unit, configured to acquire a preset slicing value; wherein the slicing value includes a first slicing value corresponding to a high-resolution image and a second slicing value corresponding to a low-resolution image;

[0191] A first processing unit, configured to perform a cropping process on the third satellite remote sensing image pair based on the block value to obtain a corresponding first image pair;

[0192] A second processing unit, configured to perform image spatial domain alignment processing on the first image pair to obtain a corresponding second image pair;

[0193] A determination unit is configured to use the second image pair as the fourth satellite remote sensing image pair.

[0194] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based remote sensing image super-resolution reconstruction method in the aforementioned implementation, and will not be repeated here.

[0195] In some optional implementations of this embodiment, the initial super-resolution reconstruction model includes a generator and a discriminator, the generator includes a convolution layer, a target number of feature extraction layers and an upsampling layer connected in sequence, and the discriminator adopts a spectrally normalized Unet structure; the training module 305 includes:

[0196] A division submodule, used to divide the remote sensing image samples into a training data set and a test data set according to a preset ratio;

[0197] A calling submodule, used for calling the initial super-resolution reconstruction model;

[0198] The fifth acquisition submodule is used to obtain a preset model training strategy;

[0199] A training submodule, configured to train the initial super-resolution reconstruction model using the training data set based on the model training strategy to obtain a corresponding designated model;

[0200] An evaluation submodule, used for performing a performance evaluation on the specified model based on the test data set;

[0201] The third determination submodule is used to use the specified model as the super-resolution reconstruction model if the specified model passes the performance evaluation.

[0202] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based remote sensing image super-resolution reconstruction method in the aforementioned implementation, and will not be repeated here.

[0203] In some optional implementations of this embodiment, the fifth acquisition submodule includes:

[0204] A second acquisition unit, used for acquiring target optimizer information;

[0205] A third acquisition unit, used to acquire a preset learning rate adjustment strategy;

[0206] A fourth acquisition unit, used to acquire a target loss function;

[0207] A generating unit is used to generate the model training strategy based on the target optimizer information, the learning rate adjustment strategy and the target loss function.

[0208] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based remote sensing image super-resolution reconstruction method in the aforementioned implementation, and will not be repeated here.

[0209] In some optional implementations of this embodiment, the remote sensing image super-resolution reconstruction device based on artificial intelligence also includes:

[0210] The second acquisition module is used to obtain a preset model deployment format;

[0211] An export module, used to export the super-resolution reconstruction model based on the model deployment format to obtain a target super-resolution reconstruction model;

[0212] A processing module is used to store and integrate the target super-resolution reconstruction model.

[0213] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based remote sensing image super-resolution reconstruction method in the aforementioned implementation, and will not be repeated here.

[0214] To solve the above technical problems, the present application also provides a computer device. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0215] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected through a system bus. It should be noted that the figure only shows a computer device 4 with components 41-43, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (Application Specific Integrated Circuit, ASIC), programmable gate arrays (Field-Programmable Gate Array, FPGA), digital processors (Digital Signal Processor, DSP), embedded devices, etc.

[0216] The computer device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may interact with a user through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device.

[0217] The memory 41 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 41 can be an internal storage unit of the computer device 4, such as a hard disk or memory of the computer device 4. In other embodiments, the memory 41 can also be an external storage device of the computer device 4, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. equipped on the computer device 4. Of course, the memory 41 can also include both the internal storage unit of the computer device 4 and its external storage device. In this embodiment, the memory 41 is generally used to store the operating system and various application software installed on the computer device 4, such as computer-readable instructions of a remote sensing image super-resolution reconstruction method based on artificial intelligence, etc. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or are to be output.

[0218] The processor 42 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments. The processor 42 is generally used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to run the computer-readable instructions stored in the memory 41 or process data, such as computer-readable instructions for running the artificial intelligence-based remote sensing image super-resolution reconstruction method.

[0219] The network interface 43 may include a wireless network interface or a wired network interface. The network interface 43 is generally used to establish a communication connection between the computer device 4 and other electronic devices.

[0220] The present application also provides another embodiment, namely, providing a computer-readable storage medium, which stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor to enable the at least one processor to perform the steps of the above-mentioned artificial intelligence-based remote sensing image super-resolution reconstruction method.

[0221] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0222] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. A remote sensing image super-resolution reconstruction method based on artificial intelligence, characterized in that: The steps include: Acquire a pair of pre-collected satellite remote sensing images of a target area; wherein the pair of satellite remote sensing images includes a first satellite remote sensing image having a first resolution and a second satellite remote sensing image having a second resolution, and the first resolution is greater than the second resolution; Performing interpolation processing on the second satellite remote sensing image to obtain a corresponding third satellite remote sensing image; wherein the resolution of the third satellite remote sensing image is a third resolution, and the third resolution is greater than the second resolution; Constructing a corresponding first satellite remote sensing image pair based on the first satellite remote sensing image and the third satellite remote sensing image; Preprocessing the first satellite remote sensing image pair to obtain corresponding remote sensing image samples; Based on the remote sensing image samples, a preset initial super-resolution reconstruction model is trained and evaluated to obtain a super-resolution reconstruction model that meets preset performance expectations; The remote sensing image to be processed is subjected to super-resolution reconstruction processing based on the super-resolution reconstruction model to obtain a corresponding target remote sensing image.

2. The method for super-resolution reconstruction of remote sensing images based on artificial intelligence according to claim 1, characterized in that: The step of performing interpolation processing on the second satellite remote sensing image to obtain a corresponding third satellite remote sensing image specifically includes: Get the target interpolation method; Obtaining a preset target resolution; wherein the target resolution is equal to the third resolution; performing interpolation processing on the second satellite remote sensing image based on the target interpolation method and the target resolution to obtain a processed second satellite remote sensing image; The processed second satellite remote sensing image is used as the third satellite remote sensing image.

3. The remote sensing image super-resolution reconstruction method based on artificial intelligence according to claim 1 is characterized in that: The step of preprocessing the first satellite remote sensing image pair to obtain corresponding remote sensing image samples specifically includes: Get the preset image processing strategy; Processing the first satellite remote sensing image pair based on the image processing strategy to obtain a corresponding second satellite remote sensing image pair; Get preset ground truth high-resolution images; Performing registration processing on the second satellite remote sensing image pair based on a preset registration strategy and the ground true high-resolution image to obtain a corresponding third satellite remote sensing image pair; Performing cropping and alignment processing on the third satellite remote sensing image pair to obtain a corresponding fourth satellite remote sensing image pair; The fourth satellite remote sensing image pair is used as the remote sensing image sample.

4. The method for super-resolution reconstruction of remote sensing images based on artificial intelligence according to claim 3 is characterized in that: The step of performing cropping and alignment processing on the third satellite remote sensing image pair to obtain a corresponding fourth satellite remote sensing image pair specifically includes: Obtaining preset slicing values; wherein the slicing values ​​include a first slicing value corresponding to a high-resolution image and a second slicing value corresponding to a low-resolution image; Performing a cropping process on the third satellite remote sensing image pair based on the block value to obtain a corresponding first image pair; Performing image spatial domain alignment processing on the first image pair to obtain a corresponding second image pair; The second image pair is used as the fourth satellite remote sensing image pair.

5. The method for super-resolution reconstruction of remote sensing images based on artificial intelligence according to claim 1, characterized in that: The initial super-resolution reconstruction model includes a generator and a discriminator, wherein the generator includes a convolution layer, a target number of feature extraction layers and an upsampling layer connected in sequence, and the discriminator adopts a spectrally normalized Unet structure; before the step of training and evaluating the preset initial super-resolution reconstruction model based on the remote sensing image sample to obtain a super-resolution reconstruction model that meets the preset performance expectations, it also includes: Dividing the remote sensing image samples into a training data set and a test data set according to a preset ratio; Calling the initial super-resolution reconstruction model; Get the preset model training strategy; Based on the model training strategy, the initial super-resolution reconstruction model is trained using the training data set to obtain a corresponding designated model; Performing a performance evaluation on the specified model based on the test data set; If the specified model passes the performance evaluation, the specified model is used as the super-resolution reconstruction model.

6. The method for super-resolution reconstruction of remote sensing images based on artificial intelligence according to claim 5, characterized in that: The step of obtaining a preset model training strategy specifically includes: Get target optimizer information; Get the preset learning rate adjustment strategy; Get the target loss function; The model training strategy is generated based on the target optimizer information, the learning rate adjustment strategy, and the target loss function.

7. The method for super-resolution reconstruction of remote sensing images based on artificial intelligence according to claim 1, characterized in that: After the step of training and evaluating the preset initial super-resolution reconstruction model based on the remote sensing image samples to obtain a super-resolution reconstruction model that meets the preset performance expectations, the method further includes: Get the preset model deployment format; Exporting the super-resolution reconstruction model based on the model deployment format to obtain a target super-resolution reconstruction model; The target super-resolution reconstruction model is stored and integrated.

8. A remote sensing image super-resolution reconstruction device based on artificial intelligence, characterized in that: include: A first acquisition module is used to acquire a pair of satellite remote sensing images of a target area collected in advance; wherein the pair of satellite remote sensing images includes a first satellite remote sensing image with a first resolution and a second satellite remote sensing image with a second resolution, and the first resolution is greater than the second resolution; An interpolation module, configured to perform interpolation processing on the second satellite remote sensing image to obtain a corresponding third satellite remote sensing image; wherein the resolution of the third satellite remote sensing image is a third resolution, and the third resolution is greater than the second resolution; A construction module, configured to construct a corresponding first satellite remote sensing image pair based on the first satellite remote sensing image and the third satellite remote sensing image; A preprocessing module, used for preprocessing the first satellite remote sensing image pair to obtain corresponding remote sensing image samples; A training module, used to train and evaluate a preset initial super-resolution reconstruction model based on the remote sensing image samples to obtain a super-resolution reconstruction model that meets preset performance expectations; The reconstruction module is used to perform super-resolution reconstruction processing on the remote sensing image to be processed based on the super-resolution reconstruction model to obtain the corresponding target remote sensing image.

9. A computer device, characterized in that: It includes a memory and a processor, wherein the memory stores computer-readable instructions, and when the processor executes the computer-readable instructions, the steps of the remote sensing image super-resolution reconstruction method based on artificial intelligence as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the artificial intelligence-based remote sensing image super-resolution reconstruction method as described in any one of claims 1 to 7.

Citation Information

Cited By

  • Water quality detection model training method, device and equipment and storage medium

    CN122473587A