Remote sensing image processing method and system

Abstract

The present invention provides a remote sensing image processing method and system, which relates to the field of image processing technology. The method includes: obtaining multi-source remote sensing image data from multiple different platforms, and respectively obtaining the image quality and information content of each type of image data; obtaining a demand label according to the remote sensing image data demand scene, and obtaining a basic weight allocation strategy based on the demand label; adjusting the basic weight allocation strategy according to the image quality and information content to obtain an optimized weight allocation strategy; performing image fusion on the multi-source remote sensing image data according to the optimized weight allocation strategy to obtain a fused remote sensing image. When performing image data fusion, the method adopts different weight ratios in different processing stages. This method can more flexibly utilize the advantages of images from different sources, and adjust the fusion strategy according to specific task requirements. It can effectively fuse remote sensing images from different platforms and improve the quality and application value of fused images.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a remote sensing image processing method and system. Background Art

[0002] With the development of remote sensing technology, remote sensing data can now be obtained from a variety of platforms, including satellites, aircraft, drones, and ground sensors. These data sources provide different spatial resolutions, spectral resolutions, and temporal resolutions. Multi-source remote sensing image data fusion is a technology that combines remote sensing images from different sensors and platforms to generate a new image that integrates multiple information. By fusing data from different sources, the spatial resolution, spectral resolution, or temporal resolution of the final image can be improved, thereby obtaining a higher quality image. Multi-source remote sensing image data fusion technology provides strong support for scientific research, environmental protection, urban management, and other fields. In practical applications, multi-source remote sensing image data fusion is a complex process. Due to certain differences in remote sensing data from different sources, the processed remote sensing image data cannot adapt well to the characteristics of different data sources, which reduces the quality and application value of the fusion results.

[0003] Therefore, how to provide a remote sensing image data processing method with better performance according to the characteristics and actual needs of different application scenarios is a problem that needs to be solved urgently. Summary of the invention

[0004] In order to improve the above problems, the present invention provides a remote sensing image processing method and system.

[0005] A first aspect of an embodiment of the present invention provides a remote sensing image processing method, the method comprising:

[0006] Acquire multi-source remote sensing image data from multiple different platforms, wherein the multi-source remote sensing data includes satellite image data, aerial image data, drone image data, and ground sensor image data, and respectively acquire image quality and information content of each type of image data;

[0007] Obtaining a demand label according to a remote sensing image data demand scenario, and obtaining a basic weight allocation strategy based on the demand label;

[0008] Adjusting the basic weight allocation strategy according to the image quality and information content to obtain an optimized weight allocation strategy;

[0009] The multi-source remote sensing image data are subjected to image fusion according to the optimized weight distribution strategy to obtain a fused remote sensing image.

[0010] Optionally, the step of adjusting the basic weight allocation strategy according to the image quality and information content to obtain an optimized weight allocation strategy specifically includes:

[0011] Adjusting the basic weight allocation strategy for the first time according to the image quality to obtain a first weight allocation strategy applied to the preprocessing stage of image fusion;

[0012] Adjusting the basic weight allocation strategy for a second time according to the information content to obtain a second weight allocation strategy applied to the image fusion feature extraction stage;

[0013] The basic weight allocation strategy is adjusted for the third time according to the image quality and information content to obtain a third weight allocation strategy applied to the data fusion stage of image fusion.

[0014] Optionally, the step of performing image fusion on the multi-source remote sensing image data based on the optimized weight allocation strategy to obtain a fused remote sensing image specifically includes:

[0015] Performing a preprocessing operation on the multi-source remote sensing image data according to the first weight distribution strategy, wherein the preprocessing operation includes geometric correction and radiation correction;

[0016] Performing a feature extraction operation on the preprocessed multi-source remote sensing image data according to the second weight distribution strategy, wherein the feature extraction operation includes multispectral feature extraction and texture feature extraction;

[0017] According to the third weight distribution strategy, a data fusion operation is performed on the multi-source remote sensing image data after feature extraction to obtain a fused remote sensing image.

[0018] Optionally, the image quality includes at least one of spatial resolution, signal-to-noise ratio, contrast, and clarity, and the step of adjusting the basic weight allocation strategy for the first time according to the image quality to obtain a first weight allocation strategy applied to the preprocessing stage of image fusion specifically includes:

[0019] Scoring the remote sensing image data from different sources according to the included items of the image quality to obtain a first scoring result;

[0020] Adjusting the basic weight allocation strategy according to the first scoring result and the association between the requirement label and the included items of the image quality;

[0021] The weight proportions of remote sensing image data from different sources in the first weight allocation strategy are respectively obtained.

[0022] Optionally, the information content includes at least one of information entropy of different bands, correlation between different bands, and feature significance, and the step of performing a second adjustment on the basic weight allocation strategy according to the information content to obtain a second weight allocation strategy applied to the image fusion feature extraction stage specifically includes:

[0023] Scoring the remote sensing image data from different sources respectively according to the included items of the information content to obtain a second scoring result;

[0024] Adjusting the basic weight allocation strategy according to the second scoring result and the correlation between the requirement label and the included items of the information content;

[0025] The weight proportions of remote sensing image data from different sources in the second weight allocation strategy are respectively obtained.

[0026] Optionally, the step of adjusting the basic weight allocation strategy for a third time according to the image quality and information content to obtain a third weight allocation strategy applied to the data fusion stage of image fusion specifically includes:

[0027] Performing a weighted summation on the first scoring result and the second scoring result according to a preset weighted ratio corresponding to the demand tag to obtain a third scoring result for remote sensing image data from different sources;

[0028] The basic weight allocation strategy is adjusted according to the third scoring result to obtain the weight proportions of remote sensing image data from different sources in the third weight allocation strategy.

[0029] Optionally, the method further comprises:

[0030] Optimizing the second weight allocation strategy according to the result of the preprocessing operation;

[0031] The third weight allocation strategy is optimized according to the result of the feature extraction operation.

[0032] Optionally, the step of performing image fusion on the multi-source remote sensing image data according to the optimized weight allocation strategy to obtain a fused remote sensing image specifically further includes:

[0033] Performing feature extraction operation on the preprocessed multi-source remote sensing image data according to the optimized second weight allocation strategy;

[0034] According to the optimized third weight distribution strategy, a data fusion operation is performed on the multi-source remote sensing image data after feature extraction to obtain a fused remote sensing image.

[0035] Optionally, the step of optimizing the second weight distribution strategy according to the result of the preprocessing operation specifically includes:

[0036] Recalculate the information content of the preprocessed multi-source remote sensing image data;

[0037] A new second weight allocation strategy is obtained according to the calculated information content.

[0038] Optionally, the step of optimizing the third weight allocation strategy according to the result of the feature extraction operation specifically includes:

[0039] Recalculate the image quality and information content of multi-source remote sensing image data after preprocessing and feature extraction;

[0040] A new third weight allocation strategy is obtained according to the calculated image quality and information content.

[0041] A second aspect of an embodiment of the present invention provides a remote sensing image processing system, the system comprising:

[0042] A data acquisition unit, which acquires multi-source remote sensing image data from multiple different platforms, wherein the multi-source remote sensing data includes satellite image data, aerial image data, drone image data, and ground sensor image data, and acquires image quality and information content of each type of image data;

[0043] A demand acquisition unit obtains demand labels according to the remote sensing image data demand scenario, and obtains a basic weight allocation strategy based on the demand labels;

[0044] A strategy optimization unit, which adjusts the basic weight allocation strategy according to the image quality and information content to obtain an optimized weight allocation strategy;

[0045] The image fusion module adjusts the basic weight allocation strategy according to the image quality and information content to obtain an optimized weight allocation strategy.

[0046] In summary, the present invention provides a remote sensing image processing method and system, which adopts different weight ratios in different processing stages when performing image data fusion. This method can more flexibly utilize the advantages of images from different sources and adjust the fusion strategy according to specific task requirements, and can effectively fuse remote sensing images from different platforms, thereby improving the quality and application value of the fused image. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0048] Figure 1 A method flow chart of a remote sensing image processing method according to an embodiment of the present invention;

[0049] Figure 2 4 is a functional module block diagram of a remote sensing image processing system according to an embodiment of the present invention.

[0050] Reference numerals:

[0051] Data acquisition unit 110; demand acquisition unit 120; strategy optimization unit 130; image fusion module unit 140. DETAILED DESCRIPTION

[0052] With the development of remote sensing technology, remote sensing data can now be obtained from a variety of platforms, including satellites, aircraft, drones, and ground sensors. These data sources provide different spatial resolutions, spectral resolutions, and temporal resolutions. Multi-source remote sensing image data fusion is a technology that combines remote sensing images from different sensors and platforms to generate a new image that integrates multiple information. By fusing data from different sources, the spatial resolution, spectral resolution, or temporal resolution of the final image can be improved, thereby obtaining a higher quality image. Multi-source remote sensing image data fusion technology provides strong support for scientific research, environmental protection, urban management, and other fields. In practical applications, multi-source remote sensing image data fusion is a complex process. Due to certain differences in remote sensing data from different sources, the processed remote sensing image data cannot adapt well to the characteristics of different data sources, which reduces the quality and application value of the fusion results.

[0053] Therefore, how to provide a remote sensing image data processing method with better performance according to the characteristics and actual needs of different application scenarios is a problem that needs to be solved urgently.

[0054] In view of this, the designers of the present invention designed a remote sensing image processing method and system.

[0055] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0056] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. 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.

[0057] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0058] In the description of the present invention, it should be noted that the terms "top", "bottom", "inside", "outside", etc. indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, or the orientation or positional relationship in which the invented product is usually placed when in use, which is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0059] In the description of the present invention, it is also necessary to explain that, unless otherwise clearly specified and limited, the terms "set", "install", "connect", and "connect" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0060] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0061] A remote sensing image processing method provided by this embodiment will be specifically described below.

[0062] See also Figure 1 , this embodiment provides a remote sensing image processing method, the method comprising:

[0063] Step S101, obtaining multi-source remote sensing image data from multiple different platforms, wherein the multi-source remote sensing data includes satellite image data, aerial image data, drone image data, and ground sensor image data, and obtaining the image quality and information content of each type of image data.

[0064] Remote sensing image data from different platforms have different characteristics due to different data collection facilities and methods. These data sources provide different spatial resolutions, spectral resolutions, and temporal resolutions. For example, satellite images can provide large-scale coverage, while drone images can provide high-resolution details.

[0065] In order to fully reflect the characteristics of remote sensing image data from different sources and obtain better results in subsequent data fusion, it is necessary to further obtain the image quality and information content of various types of image data from different platforms.

[0066] Among them, image quality includes at least one of spatial resolution, signal-to-noise ratio, contrast, and clarity. Spatial resolution reflects the spatial detail of the image, signal-to-noise ratio reflects the ratio of signal strength to noise strength, contrast reflects the brightness difference between different areas in the image, and clarity reflects the sharpness and edge clarity of the image. Information content includes at least one of information entropy of different bands, correlation between different bands, and feature significance. Information entropy is used to measure the richness of image information, correlation reflects the correlation between different bands, and feature significance reflects the significance of specific objects or features in the image.

[0067] In actual applications, the requirements for image quality and information content are different in different application scenarios. The image quality and information content of various types of image data from different platforms are also different. Therefore, it is necessary to determine how to fuse various types of image data from different platforms based on actual needs to obtain fused image data that matches current needs.

[0068] Specifically, there are significant differences in image quality and information content between various types of image data from different platforms. These differences are mainly reflected in the following aspects:

[0069] 1. Image quality;

[0070] a. Spatial resolution;

[0071] Satellite imagery data: typically have low to medium spatial resolution (e.g., 30 m resolution for Landsat 8, 0.31 m resolution for WorldView-4).

[0072] Aerial imagery data: The spatial resolution is relatively high, typically between a few centimeters and tens of centimeters, depending on the flight altitude and sensor type.

[0073] UAV image data: can provide very high spatial resolution, ranging from a few millimeters to a few centimeters, which is very suitable for local high-precision applications.

[0074] Ground sensor image data: The spatial resolution is very high, reaching the millimeter level, which is suitable for very fine measurement and monitoring.

[0075] b. Temporal resolution;

[0076] Satellite imagery data: Temporal resolution depends on the satellite revisit period, ranging from days to weeks.

[0077] Aerial image data: It has a high temporal resolution and can be flexibly arranged according to needs, but is subject to weather and flight conditions.

[0078] Drone image data: Very high temporal resolution, allowing for frequent data collection as needed.

[0079] Ground sensor image data: The temporal resolution is very high, which allows continuous monitoring and is suitable for real-time or near real-time applications.

[0080] c. Signal-to-noise ratio;

[0081] Satellite image data: The signal-to-noise ratio is usually low because the signal travels a long distance and is greatly affected by the atmosphere.

[0082] Aerial image data: The signal-to-noise ratio is higher because the flight altitude is low and the signal transmission distance is short.

[0083] Drone image data: The signal-to-noise ratio is very high because the flight altitude is lower and the signal transmission distance is shorter.

[0084] Ground sensor image data: The signal-to-noise ratio is very high because the sensor is in direct contact with or close to the target.

[0085] d. Contrast and clarity;

[0086] Satellite image data: Contrast and clarity are greatly affected by atmospheric conditions and sensor performance.

[0087] Aerial image data: has high contrast and clarity, but is affected by flight altitude and weather conditions.

[0088] Drone image data: Very high contrast and clarity, less affected by weather conditions.

[0089] Ground sensor image data: Very high contrast and clarity, suitable for close-range observations.

[0090] 2. Information content;

[0091] a. Spectral resolution;

[0092] Satellite imagery data: Multispectral and hyperspectral sensors can provide rich spectral information, covering visible to infrared bands.

[0093] Aerial imagery data: Multispectral and hyperspectral sensors also provide rich spectral information, but typically with higher spectral resolution.

[0094] UAV image data: Multispectral and hyperspectral sensors can provide high spectral resolution and are suitable for local high-precision applications.

[0095] Ground sensor image data: can provide very detailed spectral information, suitable for detailed analysis of specific ground objects.

[0096] b. Feature prominence;

[0097] Satellite image data: suitable for large-scale object recognition and classification, with low feature significance.

[0098] Aerial image data: suitable for medium-range object recognition and classification, with high feature significance.

[0099] UAV image data: suitable for local high-precision object recognition and classification, with very high feature significance.

[0100] Ground sensor image data: suitable for very fine object recognition and classification, with very high feature significance.

[0101] 3. Application scenarios;

[0102] Satellite image data: suitable for global or regional scale environmental monitoring, land use change detection, disaster assessment, etc.

[0103] Aerial image data: suitable for urban planning, agricultural monitoring, forest management, infrastructure construction, etc.

[0104] UAV image data: suitable for high-precision monitoring of small areas, such as farmland management, construction site monitoring, archaeological surveys, etc.

[0105] Ground sensor image data: suitable for detailed monitoring of specific locations, such as soil moisture monitoring, water quality monitoring, vegetation health monitoring, etc.

[0106] Step S102, obtaining a demand label according to a remote sensing image data demand scenario, and obtaining a basic weight allocation strategy based on the demand label.

[0107] There are many application scenarios for remote sensing image data, such as land cover classification, urban change detection, disaster monitoring, agricultural monitoring, forest fire monitoring, ocean monitoring, etc. Different demand scenarios have different requirements for the use of remote sensing image data. The requirements of different demand scenarios can be classified in advance to obtain the demand labels under different demand scenarios, and these demand labels can be matched with the data parameter items and numerical requirements of remote sensing image data, including image quality, information content, sensor characteristics, time correlation, spatial consistency, etc.

[0108] When this method is actually implemented, the demand label is first obtained based on the demand scenario, and then the corresponding relationship between the demand label and the data parameter items of the remote sensing image data is used to determine which parameter items are more important under the current demand, and then the weights are determined according to the importance of these parameter items. Finally, based on the data characteristics of various image data from different platforms, that is, which platform has a higher accuracy or more comprehensive data volume in the parameter items with a large weight in the remote sensing data, the basic weight allocation strategy for various image data from different platforms can be obtained.

[0109] To give a specific example, assume that the application scenario is: urban land use classification.

[0110] In this scenario, high spatial resolution is required to identify buildings, roads, and other urban objects. Color fidelity is also important to distinguish different types of surface cover. Multispectral information is also needed to distinguish different types of objects (such as vegetation, water bodies, buildings, etc.). The near-infrared band is very important for vegetation health monitoring and distinguishing vegetation from non-vegetated areas.

[0111] Based on the above requirements, remote sensing data from different sources are used, among which drone images and ground sensor data have high spatial resolution, detailed spectral information and texture details, and satellite images and aerial images have large coverage. In the basic weight allocation strategy in this scenario, ground sensor data and drone images have higher weights, while satellite images and aerial images have relatively lower weights.

[0112] It should be noted that the basic weight allocation strategy is mainly based on the demand label of the demand scenario. It is also required to make appropriate adjustments based on the image quality and information content of each type of remote sensing data obtained to obtain a more accurate weight allocation strategy.

[0113] Step S103, adjusting the basic weight allocation strategy according to the image quality and information content to obtain an optimized weight allocation strategy.

[0114] The remote sensing data from different platforms are evaluated by image quality and information content. The weight of the data with better performance can be appropriately increased to optimize the weight allocation strategy.

[0115] The process of multi-source remote sensing image data fusion mainly includes the following steps:

[0116] 1. Collect data;

[0117] Acquire remote sensing images from different platforms (e.g. satellites, drones, ground sensors).

[0118] 2. Preprocessing stage;

[0119] Perform geometric correction, radiation correction and time synchronization. Geometric correction ensures that all images are aligned in space, radiation correction eliminates radiation differences caused by atmospheric effects and other factors, and time synchronization is used to reduce the impact of time differences.

[0120] 3. Feature extraction stage;

[0121] Extract multi-spectral features and texture features.

[0122] 4. Image fusion stage;

[0123] Select appropriate fusion method for fusion;

[0124] 5. Post-processing stage;

[0125] Adjust color balance based on the fusion results and enhance image details through edge-preserving filters and other techniques.

[0126] The above three stages 2-4 all need to be based on the optimization weight allocation strategy, and perform corresponding operations according to the specific weights allocated to the remote sensing data from different platforms. Considering the different requirements for remote sensing data at different stages, weight allocation can be performed for each stage separately. Specifically, as a preferred method of an embodiment of the present invention, step S103 specifically includes:

[0127] Adjusting the basic weight allocation strategy for the first time according to the image quality to obtain a first weight allocation strategy applied to the preprocessing stage of image fusion;

[0128] Adjusting the basic weight allocation strategy for a second time according to the information content to obtain a second weight allocation strategy applied to the image fusion feature extraction stage;

[0129] The basic weight allocation strategy is adjusted for the third time according to the image quality and information content to obtain a third weight allocation strategy applied to the data fusion stage of image fusion.

[0130] For each stage that needs to allocate weights, appropriate optimization can be performed on the basis of the basic weight allocation strategy according to the image quality and information content of remote sensing data from different platforms. It can be seen that the basic weight allocation strategy determines the overall direction of the weight allocation strategy, and through specific optimization, more suitable optimization can be performed for different needs at different stages. For each stage, the specific optimization method is as follows:

[0131] As a preferred embodiment of the present invention, the calculation method of the first weight allocation strategy specifically includes:

[0132] Scoring the remote sensing image data from different sources according to the included items of the image quality to obtain a first scoring result;

[0133] Adjusting the basic weight allocation strategy according to the first scoring result and the association between the requirement label and the included items of the image quality;

[0134] The weight proportions of remote sensing image data from different sources in the first weight allocation strategy are respectively obtained.

[0135] As a preferred embodiment of the present invention, the calculation method of the second weight distribution strategy specifically includes:

[0136] Scoring the remote sensing image data from different sources respectively according to the included items of the information content to obtain a second scoring result;

[0137] Adjusting the basic weight allocation strategy according to the second scoring result and the correlation between the requirement label and the included items of the information content;

[0138] The weight proportions of remote sensing image data from different sources in the second weight allocation strategy are respectively obtained.

[0139] As a preferred embodiment of the present invention, the calculation method of the third weight allocation strategy specifically includes:

[0140] Performing a weighted summation on the first scoring result and the second scoring result according to a preset weighted ratio corresponding to the demand tag to obtain a third scoring result for remote sensing image data from different sources;

[0141] The basic weight allocation strategy is adjusted according to the third scoring result to obtain the weight proportions of remote sensing image data from different sources in the third weight allocation strategy.

[0142] In order to better systematically analyze the specific conditions of remote sensing data from different platforms, in this embodiment, a quantitative scoring method is adopted, and a set of scoring criteria are predefined to measure image quality and information content. The specific evaluation criteria include:

[0143] Image Quality:

[0144] Spatial resolution: The degree of spatial detail in an image. High-resolution images generally provide more detail information.

[0145] Signal-to-Noise Ratio: The ratio of signal strength to noise strength. Images with high SNR are generally of better quality.

[0146] Contrast: The difference in brightness between different areas in an image. High contrast helps distinguish different objects.

[0147] Clarity: The sharpness and edge definition of an image. High-definition images make it easier to identify the boundaries of objects.

[0148] Information content:

[0149] Information entropy: measures the richness of image information. A high information entropy means the image contains more information.

[0150] Correlation: The correlation between different bands. Low correlation means that the information provided by the bands is more independent and complementary.

[0151] Feature saliency: The degree to which a particular object or feature is prominent in an image. High saliency means that the object in the image is easier to identify.

[0152] Each item corresponds to a specific score. The scores of each item are calculated for remote sensing data from different platforms. The image quality and information content are accumulated and summed to obtain the first scoring result and the second scoring result. On this basis, the first scoring result and the second scoring result are weighted and summed according to the preset weighted ratio corresponding to the demand label to obtain the third scoring result. The determination of the weighted ratio is related to the demand label. Under different demand labels, the weighted ratios of the first scoring result and the second scoring result are different when weighted summing.

[0153] In the preprocessing stage, higher spatial resolution, signal-to-noise ratio and clarity are required. In the feature extraction stage, bands with rich information and larger entropy values ​​as well as independent (low correlation) and important bands are required. In the data fusion stage, the required range coverage and specific content of the image are determined based on the required labels.

[0154] At different stages, according to the actual scoring results, corresponding adjustments are made on the basis of the basic weight allocation strategy. For example, in the preprocessing stage, the satellite image data has the highest weight in the basic weight allocation strategy, but its first scoring result has a low score, so in the first weight allocation strategy, the weight is correspondingly lowered.

[0155] Through the above process, weight allocation strategies for three operation stages that require specific weights to be executed are obtained respectively, namely, the first weight allocation, the second weight allocation and the third weight allocation.

[0156] Step S104, performing image fusion on the multi-source remote sensing image data according to the optimized weight distribution strategy to obtain a fused remote sensing image.

[0157] As a preferred embodiment of the present invention, step S104 specifically includes:

[0158] Performing a preprocessing operation on the multi-source remote sensing image data according to the first weight distribution strategy, wherein the preprocessing operation includes geometric correction and radiation correction;

[0159] Performing a feature extraction operation on the preprocessed multi-source remote sensing image data according to the second weight distribution strategy, wherein the feature extraction operation includes multispectral feature extraction and texture feature extraction;

[0160] According to the third weight distribution strategy, a data fusion operation is performed on the multi-source remote sensing image data after feature extraction to obtain a fused remote sensing image.

[0161] According to the obtained optimized weight allocation strategy, according to the actual execution stage, the corresponding weight allocation strategy is used to perform specific operations, and finally a fused remote sensing image is obtained that matches the current scene requirements and gives full play to the characteristics of remote sensing image data from different sources.

[0162] Image fusion operations can use common image fusion methods, including brightness-hue-saturation (IHS) transform, principal component analysis (PCA) transform, Schmidt orthogonalization (Gram-Schmidt, GS) transform, wavelet transform, high-pass filtering variation, etc. The specific method selection may depend on the specific application scenario requirements and the platform type of the acquired remote sensing image. For example, when it is necessary to maintain good color fidelity, IHS transform may be given priority; when emphasizing detail preservation and noise reduction, wavelet transform or deep learning-based methods may be preferred.

[0163] The operations of the above three stages, after being specifically performed, will have a certain impact on the performance of subsequent stages. Therefore, as a preferred implementation, the method further includes:

[0164] Optimizing the second weight allocation strategy according to the result of the preprocessing operation;

[0165] The third weight allocation strategy is optimized according to the result of the feature extraction operation.

[0166] Specifically, the information content of the preprocessed multi-source remote sensing image data is recalculated;

[0167] A new second weight allocation strategy is obtained according to the calculated information content.

[0168] Recalculate the image quality and information content of multi-source remote sensing image data after preprocessing and feature extraction;

[0169] A new third weight allocation strategy is obtained according to the calculated image quality and information content.

[0170] In actual applications, it is found that when executing step S101, the image quality and information content obtained are not necessarily fixed, and after being processed by certain operations, the image quality and information content may change. Therefore, in order to improve the matching degree of the optimization weight allocation strategy in each stage, after each stage of operation is completed, the image quality and information content are recalculated, and then the optimization weight allocation strategy for the subsequent stage is obtained based on the recalculated results.

[0171] For example, in the geometric correction stage, if it is found that a certain type of remote sensing image has a large error after geometric correction, the weight of the image can be appropriately reduced. In the radiation correction stage, if the signal-to-noise ratio of a certain type of remote sensing image is significantly improved after radiation correction, the weight of the image can be appropriately increased. In the feature extraction stage, if the feature significance of a certain band is high, the weight of the band can be increased.

[0172] At this time, as a preferred method of the embodiment of the present invention, step S104 specifically includes:

[0173] Performing feature extraction operation on the preprocessed multi-source remote sensing image data according to the optimized second weight allocation strategy;

[0174] According to the optimized third weight distribution strategy, a data fusion operation is performed on the multi-source remote sensing image data after feature extraction to obtain a fused remote sensing image.

[0175] If the second weight allocation strategy or the third weight allocation strategy is optimized again based on the execution results of each stage, when executing the steps of the subsequent stages, it is necessary to perform operations of the corresponding stages according to the optimized second weight allocation strategy or the third weight allocation strategy.

[0176] On the basis of the above scheme, in order to further improve the quality of fused remote sensing images, as a preferred implementation method, when calculating the weight distribution strategy, the image can be divided into different areas (such as building-dense areas, vegetation-covered areas, water areas, etc.), and the image quality and information content are evaluated in each area. Then, according to the above method, the weight is adjusted according to the evaluation results of each area. For example, if it is found in the preprocessing stage that a certain area is covered with more vegetation, the weight of the near-infrared band is increased in this area to enhance the expression of vegetation information.

[0177] On the basis of the above scheme, in order to further improve the quality of fused remote sensing images, a feedback mechanism can be established: Introduce a feedback mechanism in the entire process to ensure that the weight adjustment of each stage is based on the results of the previous stage. Finally, according to the results of post-processing and evaluation, fine-tune the weights of each stage. For example, if the importance of a certain band is underestimated in the feature extraction stage, the weight of the band can be appropriately increased in the subsequent fusion stage.

[0178] The remote sensing image processing method provided by the present invention is described below through a specific scenario.

[0179] Scenario 1: Farmland management;

[0180] Demand focus;

[0181] Image quality: High spatial resolution is required to identify crop row spacing, field roads, and irrigation systems. Color fidelity is also important to distinguish different crop types and health conditions.

[0182] Information content: Multispectral information is needed to monitor crop growth status and health. The near-infrared band is particularly important for vegetation index calculations such as NDVI.

[0183] Weight selection;

[0184] 1. Preprocessing;

[0185] Geometric correction:

[0186] Satellite imagery: Initial weight 0.3, suitable for large area coverage, but may require larger geometric corrections.

[0187] Aerial imagery: initial weight 0.5, higher spatial resolution and better geometric accuracy.

[0188] UAV image: initial weight 0.7, very high spatial resolution, suitable for local high-precision correction.

[0189] Ground sensor data: initial weight 0.9, very high spatial resolution, suitable for detailed corrections at specific locations.

[0190] Radiometric correction:

[0191] Satellite imagery: Initial weight is 0.4, which is greatly affected by the atmosphere.

[0192] Aerial imagery: Initial weight is 0.6, which is less affected by the atmosphere.

[0193] Drone image: Initial weight 0.8, least affected by the atmosphere.

[0194] Ground sensor data: initial weight 0.9, not affected by the atmosphere.

[0195] 2. Feature extraction;

[0196] Multispectral features;

[0197] Satellite imagery: initial weight 0.5, rich multispectral information.

[0198] Aerial imagery: initial weight 0.7, higher signal-to-noise ratio and multispectral information.

[0199] UAV imagery: initial weight 0.8, high spatial resolution and multispectral information.

[0200] Ground sensor data: initial weight 0.9, very detailed spectral information.

[0201] Texture features:

[0202] Satellite imagery: Initial weight 0.3, less texture detail.

[0203] Aerial imagery: Initial weight 0.5, moderate texture detail.

[0204] Drone image: initial weight 0.7, rich texture details.

[0205] Ground sensor data: initial weight 0.9, very rich texture details.

[0206] 3. Data fusion;

[0207] Initial Fusion:

[0208] Satellite imagery: initial weight 0.3, wide coverage.

[0209] Aerial imagery: Initial weight 0.5, medium range coverage and high spatial resolution.

[0210] Drone image: initial weight 0.7, local high-resolution details.

[0211] Ground sensor data: initial weight 0.9, very fine data.

[0212] Content-based weighting:

[0213] Increase the weight of drone imagery in crop-growing areas (e.g. 0.8).

[0214] Increase the weight of the near infrared band (for example, 0.7) in vegetated areas.

[0215] In this scenario, high spatial resolution and multispectral information, especially near-infrared bands, are relied upon, so UAV images and ground sensor data should also be given a high weight at all stages.

[0216] Scenario 2: Large-scale forest cover monitoring;

[0217] Demand focus;

[0218] Image Quality: Medium to high spatial resolution is required to identify forest cover types (e.g. coniferous, broadleaf, etc.). Color fidelity is also important to distinguish between different types of vegetation.

[0219] Information content: Multispectral information is needed to monitor forest health and trends. Near-infrared bands are particularly important for vegetation index calculations such as NDVI. At the same time, large-scale coverage is required to monitor forest changes across an entire region.

[0220] Weight selection;

[0221] 1. Preprocessing;

[0222] Geometric correction:

[0223] Satellite imagery: Initial weight 0.9, suitable for large-area coverage and with high geometric accuracy.

[0224] Aerial imagery: initial weight 0.6, higher spatial resolution and better geometric accuracy.

[0225] Drone imagery: Initial weight 0.4, very high spatial resolution, but limited coverage.

[0226] Ground sensor data: initial weight 0.2, extremely high spatial resolution, but only applicable to small local areas.

[0227] Radiometric correction:

[0228] Satellite images: The initial weight is 0.8. Although they are greatly affected by the atmosphere, they can be compensated by the atmospheric correction algorithm.

[0229] Aerial imagery: Initial weight is 0.6, which is less affected by the atmosphere.

[0230] Drone image: Initial weight 0.5, least affected by the atmosphere.

[0231] Ground sensor data: initial weight 0.3, not affected by the atmosphere, but limited coverage.

[0232] 2. Feature extraction;

[0233] Multispectral features:

[0234] Satellite imagery: Initial weight 0.8, rich multispectral information, suitable for large-scale monitoring.

[0235] Aerial imagery: initial weight 0.6, higher signal-to-noise ratio and multispectral information.

[0236] UAV imagery: initial weight 0.5, high spatial resolution and multispectral information, but limited coverage.

[0237] Ground sensor data: initial weight 0.3, very detailed spectral information, but only applicable to a small local area.

[0238] Texture features:

[0239] Satellite imagery: Initial weight 0.7, moderate texture detail, suitable for large-scale monitoring.

[0240] Aerial imagery: Initial weight 0.5, moderate texture detail.

[0241] Drone imagery: Initial weight 0.4, rich texture details but limited coverage.

[0242] Ground sensor data: Initial weight 0.2, very rich texture details, but only suitable for small local areas.

[0243] 3. Data fusion;

[0244] Initial Fusion:

[0245] Satellite imagery: initial weight 0.9, wide coverage.

[0246] Aerial imagery: Initial weight 0.6, medium range coverage and high spatial resolution.

[0247] Drone image: initial weight 0.4, local high-resolution details.

[0248] Ground sensor data: Initial weight 0.2, very fine data, but limited coverage.

[0249] Content-based weighting:

[0250] Increase the weight of satellite images (e.g. 0.9) over areas with large forest cover.

[0251] Increase the weight of the near infrared band (for example, 0.8) in areas where specific forest types or health conditions are monitored.

[0252] In the application scenario of large-scale forest cover monitoring, the weight allocation of each step takes into account the requirements of image quality and information content. Specifically:

[0253] Preprocessing stage:

[0254] Geometric correction: Satellite imagery is given the highest weight due to its wide coverage and high geometric accuracy.

[0255] Radiometric correction: Satellite images are also weighted higher, although they are more affected by the atmosphere, but this can be compensated by correction algorithms.

[0256] Feature extraction stage:

[0257] Multispectral features: Satellite images have the highest weight because they provide rich multispectral information and are suitable for large-scale monitoring.

[0258] Texture features: Satellite images are also weighted higher and are suitable for large-scale texture analysis.

[0259] Data fusion stage:

[0260] Initial fusion: Satellite imagery is given the highest weight because it provides large area coverage.

[0261] Content-based weighting: Increase the weight of satellite imagery over areas with large forest cover, and increase the weight of near-infrared bands over areas where specific forest types or health conditions are monitored.

[0262] This phased weight allocation strategy ensures that each stage fully utilizes the advantages of satellite imagery, thereby achieving efficient large-scale forest cover monitoring.

[0263] In summary, the remote sensing image processing method provided by the embodiment of the present invention adopts different weight ratios in different processing stages when performing image data fusion. This method can more flexibly utilize the advantages of images from different sources, and adjust the fusion strategy according to specific task requirements, and can effectively fuse remote sensing images from different platforms, thereby improving the quality and application value of the fused image. Through the above method, it can be ensured that the image quality and information content are comprehensively considered at each stage of image fusion, and the weights are dynamically adjusted, thereby improving the quality and application value of the final fused image. This method can not only make full use of the advantages of images from different sources, but also show stronger adaptability and robustness in different application scenarios.

[0264] like Figure 2 As shown, an embodiment of the present invention further provides a remote sensing image processing system, the system comprising:

[0265] The data acquisition unit 110 acquires multi-source remote sensing image data from multiple different platforms, wherein the multi-source remote sensing data includes satellite image data, aerial image data, drone image data, and ground sensor image data, and acquires image quality and information content of each type of image data;

[0266] A demand acquisition unit 120 obtains a demand label according to a demand scenario of remote sensing image data, and obtains a basic weight allocation strategy based on the demand label;

[0267] A strategy optimization unit 130 adjusts the basic weight allocation strategy according to the image quality and information content to obtain an optimized weight allocation strategy;

[0268] The image fusion module 140 adjusts the basic weight allocation strategy according to the image quality and information content to obtain an optimized weight allocation strategy.

[0269] The remote sensing image processing system provided in the embodiment of the present invention is used to implement the above remote sensing image processing method, so the specific implementation method is the same as the above method and will not be repeated here.

[0270] In several embodiments disclosed in the present application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the devices, methods and computer program products according to multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of the code, and a module, a program segment or a part of the code contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0271] In addition, the functional modules in the various embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

[0272] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

Claims

1. A remote sensing image processing method, characterized in that: The method comprises: Acquire multi-source remote sensing image data from multiple different platforms, wherein the multi-source remote sensing data includes satellite image data, aerial image data, drone image data, and ground sensor image data, and respectively acquire image quality and information content of each type of image data; Obtaining a demand label according to a remote sensing image data demand scenario, and obtaining a basic weight allocation strategy based on the demand label; Adjusting the basic weight allocation strategy according to the image quality and information content to obtain an optimized weight allocation strategy; Performing image fusion on the multi-source remote sensing image data according to the optimized weight distribution strategy to obtain a fused remote sensing image; The step of adjusting the basic weight allocation strategy according to the image quality and information content to obtain an optimized weight allocation strategy specifically includes: Adjusting the basic weight allocation strategy for the first time according to the image quality to obtain a first weight allocation strategy applied to the preprocessing stage of image fusion; Adjusting the basic weight allocation strategy for a second time according to the information content to obtain a second weight allocation strategy applied to the image fusion feature extraction stage; The basic weight allocation strategy is adjusted for the third time according to the image quality and information content to obtain a third weight allocation strategy applied to the data fusion stage of image fusion.

2. The remote sensing image processing method according to claim 1, characterized in that: The step of performing image fusion on the multi-source remote sensing image data according to the optimized weight distribution strategy to obtain a fused remote sensing image specifically includes: Performing a preprocessing operation on the multi-source remote sensing image data according to the first weight distribution strategy, wherein the preprocessing operation includes geometric correction and radiation correction; Performing a feature extraction operation on the preprocessed multi-source remote sensing image data according to the second weight distribution strategy, wherein the feature extraction operation includes multispectral feature extraction and texture feature extraction; According to the third weight distribution strategy, a data fusion operation is performed on the multi-source remote sensing image data after feature extraction to obtain a fused remote sensing image.

3. The remote sensing image processing method according to claim 2, characterized in that: The image quality includes at least one of spatial resolution, signal-to-noise ratio, contrast, and clarity. The step of first adjusting the basic weight allocation strategy according to the image quality to obtain a first weight allocation strategy applied to the preprocessing stage of image fusion specifically includes: Scoring the remote sensing image data from different sources according to the included items of the image quality to obtain a first scoring result; Adjusting the basic weight allocation strategy according to the first scoring result and the association between the requirement label and the included items of the image quality; The weight proportions of remote sensing image data from different sources in the first weight allocation strategy are respectively obtained.

4. The remote sensing image processing method according to claim 3, characterized in that: The information content includes at least one of information entropy of different bands, correlation between different bands, and feature significance. The step of performing a second adjustment on the basic weight allocation strategy according to the information content to obtain a second weight allocation strategy applied to the image fusion feature extraction stage specifically includes: Scoring the remote sensing image data from different sources respectively according to the included items of the information content to obtain a second scoring result; Adjusting the basic weight allocation strategy according to the second scoring result and the correlation between the requirement label and the included items of the information content; The weight proportions of remote sensing image data from different sources in the second weight allocation strategy are respectively obtained.

5. The remote sensing image processing method according to claim 4, characterized in that: The step of adjusting the basic weight allocation strategy for the third time according to the image quality and information content to obtain a third weight allocation strategy applied to the data fusion stage of image fusion specifically includes: Performing a weighted summation on the first scoring result and the second scoring result according to a preset weighted ratio corresponding to the demand tag to obtain a third scoring result for remote sensing image data from different sources; The basic weight allocation strategy is adjusted according to the third scoring result to obtain the weight proportions of remote sensing image data from different sources in the third weight allocation strategy.

6. The remote sensing image processing method according to any one of claims 2 to 5, characterized in that: The method further comprises: Optimizing the second weight allocation strategy according to the result of the preprocessing operation; The third weight allocation strategy is optimized according to the result of the feature extraction operation.

7. The remote sensing image processing method according to claim 6, characterized in that: The step of performing image fusion on the multi-source remote sensing image data according to the optimized weight distribution strategy to obtain a fused remote sensing image specifically includes: Performing feature extraction operation on the preprocessed multi-source remote sensing image data according to the optimized second weight allocation strategy; According to the optimized third weight distribution strategy, a data fusion operation is performed on the multi-source remote sensing image data after feature extraction to obtain a fused remote sensing image.

8. The remote sensing image processing method according to claim 7, characterized in that: The step of optimizing the second weight allocation strategy according to the result of the preprocessing operation specifically includes: Recalculate the information content of the preprocessed multi-source remote sensing image data; A new second weight allocation strategy is obtained according to the calculated information content.

9. A remote sensing image processing system, characterized in that: The system comprises: A data acquisition unit, which acquires multi-source remote sensing image data from multiple different platforms, wherein the multi-source remote sensing data includes satellite image data, aerial image data, drone image data, and ground sensor image data, and acquires image quality and information content of each type of image data; A demand acquisition unit obtains demand labels according to the remote sensing image data demand scenario, and obtains a basic weight allocation strategy based on the demand labels; A strategy optimization unit, which adjusts the basic weight allocation strategy according to the image quality and information content to obtain an optimized weight allocation strategy; An image fusion module unit performs image fusion on the multi-source remote sensing image data according to the optimized weight allocation strategy to obtain a fused remote sensing image; The strategy optimization unit is specifically used for: Adjusting the basic weight allocation strategy for the first time according to the image quality to obtain a first weight allocation strategy applied to the preprocessing stage of image fusion; Adjusting the basic weight allocation strategy for a second time according to the information content to obtain a second weight allocation strategy applied to the image fusion feature extraction stage; The basic weight allocation strategy is adjusted for the third time according to the image quality and information content to obtain a third weight allocation strategy applied to the data fusion stage of image fusion.

Images